Abstract
Emotion recognition through computational modeling and analysis of physiological signals has been widely investigated in the last decade. Most of the proposed emotion recognition systems require relatively longtime series of multivariate records and do not provide accurate realtime characterizations using shorttime series. To overcome these limitations, we propose a novel personalized probabilistic framework able to characterize the emotional state of a subject through the analysis of heartbeat dynamics exclusively. The study includes thirty subjects presented with a set of standardized images gathered from the international affective picture system, alternating levels of arousal and valence. Due to the intrinsic nonlinearity and nonstationarity of the RR interval series, a specific pointprocess model was devised for instantaneous identification considering autoregressive nonlinearities up to the thirdorder according to the WienerVolterra representation, thus tracking very fast stimulusresponse changes. Features from the instantaneous spectrum and bispectrum, as well as the dominant Lyapunov exponent, were extracted and considered as input features to a support vector machine for classification. Results, estimating emotions each 10 seconds, achieve an overall accuracy in recognizing four emotional states based on the circumplex model of affect of 79.29%, with 79.15% on the valence axis, and 83.55% on the arousal axis.
Introduction
The detection and recognition of emotional information is an important topic in the field of affective computing, i.e. the study of human affects by technological systems and devices^{1}. Changes in emotional states often reflect facial, vocal, and gestural modifications in order to communicate, sometimes subunconsciously, personal feelings to other people. Such changes can be generalized across cultures, e.g. nonverbal emotional, or can be culturespecific^{2}. Since mood alteration strongly affects the normal emotional process, emotion recognition is also an ambitious objective in the field of mood disorder psychopathology. In the last decade, several efforts have tried to obtain a reliable methodology to automatically identify the emotional/mood state of a subject, starting from the analysis of facial expressions, behavioral correlates, and physiological signals. Despite such efforts, current practices still use simple mood questionnaires or interviews for emotional assessment. In mental care, for instance, the diagnosis of pathological emotional fluctuations is mainly made through the physician's experience. Several epidemiological studies report that more than two million Americans have been diagnosed with bipolar disorder^{3}, and about 82.7 million of the adult European population from 18 to 65 years of age, have been affected by at least one mental disorder^{4}. Several computational methods for emotion recognition based on variables associated with the Central Nervous System (CNS), for example the Electroencephalogram (EEG), have been recently proposed^{5,6,7,8,9,10,11,12}. These methods are justified by the fact that human emotions originate in the cerebral cortex involving several areas for their regulation and feeling. The prefrontal cortex and amygdala, in fact, represent the essence of two specific pathways: affective elicitations longer than 6 seconds allow the prefrontal cortex to encode the stimulus information and transmit it to other areas of the Central Autonomic Network (CAN) to the brainstem, thus producing a context appropriate response^{13}; briefly presented stimuli access the fast route of emotion recognition via the amygdala. Of note, it has been found that the visual cortex is involved in emotional reactions to different classes of stimuli^{14}. Dysfunctions on these CNS recruitment circuits lead to pathological effects^{15} such as anhedonia, i.e. the loss of pleasure or interest in previously rewarding stimuli, which is a core feature of major depression and other serious mood disorders.
Given the CAN involvement in emotional responses, an important direction for affective computing studies is related to changes of the Autonomic Nervous System (ANS) activity as elicited by specific emotional states. Monitoring physiological variables linked to ANS activity, in fact, can be easily performed through wearable systems, e.g. sensorized tshirts^{16,17} or gloves^{18,46}. Its dynamics is thought to be less sensitive to artifacting events than in the EEG case. Moreover, the human vagus nerve is anatomically linked to the cranial nerves that regulate social engagement via facial expression and vocalization. Engineering approaches to assess ANS patterns related to emotions constitute a relevant part of the stateoftheart methods used in affective computing. For example, a recent review written by Calvo et al.^{19} reports on emotion theories as well as on affect detection systems using physiological and speech signals (also reviewed in^{20}), face expression and movement analysis. Long multivariate recordings are currently needed to accurately characterize the emotional state of a subject. Such a constraint surely reduces the potential wide spectrum of real applications due to computational cost and number of sensors. More recently, ECG morphological analysis by HilbertHuang transform^{21}, mutual information analysis of respiratory signals^{22}, and a multiparametric approach related to ANS activity^{23} have been proposed to assess human affective states.
Experimental evidence over the past two decades shows that Heart Rate Variability (HRV) analysis, in both time and frequency domain, can provide a unique, noninvasive assessment of autonomic function^{24,25,88}. Nevertheless, HRV analysis by means of standard procedures presents several limitations when high time and frequency resolutions are needed, due mainly to associated inherent assumptions of stationarity required to define most of the relevant HRV time and frequency domain indices^{24,25}. More importantly, standard methods are generally not suitable to provide accurate nonlinear measures in the absence of information regarding phase space fitting. It has been wellaccepted by the scientific community that physiological models should be nonlinear in order to thoroughly describe the characteristics of such complex systems. Within the cardiovascular system, the complex and nonstationary dynamics of heartbeat variations have been associated to nonlinear neural interactions and integrations occurring at the neuron and receptor levels, so that the sinoatrial node responds in a nonlinear way to the changing levels of efferent autonomic inputs^{27}. In fact, HRV nonlinear measures have been demonstrated to be of prognostic value in aging and diseases^{24,25,26,28,29,30,31,32,33,34,35,36,41}. In several previous works^{37,38,39,40,41,42,43}, we have demonstrated how it is possible to estimate heartbeat dynamics in cardiovascular recordings under nonstationary conditions by means of the analysis of the probabilistic generative mechanism of the heartbeat. Concerning emotion recognition, we recently demonstrated the important role of nonlinear dynamics for a correct arousal and valence recognition from ANS signals^{44,45,46,71} including a preliminary feasibility study on the dataset considered here^{47}.
In the light of all these issues, we here propose a new methodology in the field of affective computing, able to recognize emotional swings (positive or negative), as well as two levels of arousal and valence (lowmedium and mediumhigh), using only one biosignal, the ECG, and able to instantaneously assess the subject's state even in shorttime events (<10 seconds). Emotions associated with a shorttime stimulus are identified through a selfreported label as well as four specific regions in the arousalvalence orthogonal dimension (see Fig. 1). The proposed methodology is fully based on a personalized probabilistic point process nonlinear model. In general, we model the probability function of the next heartbeat given the past Revents. The probability function is fully parametrized, considering up to cubic autoregressive WienerVolterra relationship to model its first order moment. All the considered features are estimated by the linear, quadratic, and cubic coefficients of the linear and nonlinear terms of such a relationship. As a consequence, our model provides the unique opportunity to take into account all the possible linear and nonlinear features, which can be estimated from the model parameters. Importantly as the probability function is defined at each moment in time, the parameter estimation is performed instantaneously, a feature not reliably accomplishable by using other more standard linear and nonlinear indices such as pNN50%, triangular index, RMSSD, Recurrence Quantification Analysis, etc.^{24,25,88}. In particular, the linear terms allow for the instantaneous spectral estimation, the quadratic terms allow for the instantaneous bispectral estimation, whereas the dominant Lyapunov exponent can be defined by considering the cubic terms. Of note, the use of higher order statistics (HOS) to estimate our features is encouraged by the fact that quantification of HRV nonlinear dynamics play a crucial role in emotion recognition systems^{44,48,71} extending the information given by spectral analysis, and providing useful information on the nonlinear frequency interactions.
Results
Experimental protocol
The recording paradigm related to this work has been previously described in^{44,48}. We adopted a common dimensional model which uses multiple dimensions to categorize emotions, the Circumplex Model of Affects (CMA)^{49}. The CMA used in our experiment takes into account two main dimensions conceptualized by the terms of valence and arousal (see Fig. 1). Valence represents how much an emotion is perceived as positive or negative, whereas arousal indicates how strongly the emotion is felt. Accordingly, we employed visual stimuli belonging to an international standardized database having a specific emotional rating expressed in terms of valence and arousal. Specifically, we chose the International Affective Picture System (IAPS)^{50}, which is one of the most frequently cited tools in the area of affective stimulation. The IAPS is a set of 944 images with emotional ratings based on several studies previously conducted where subjects were requested to rank these images using the self assessment manikin (both valence and arousal scales range from 0 to 10). A general overview of the experimental protocol and analysis is shown in Fig. 2. The passive affective elicitation performed through the IAPS images stimulates several cortical areas also allowing the prefrontal cortex modulation generating cognitive perceptions^{13}. An homogeneous population of 30 healthy subjects (aged from 21 to 24), not suffering from both cardiovascular and evident mental pathologies, was recruited to participate in the experiment. The experimental protocol for this study was approved by the ethical committee of the University of Pisa and an informed consent was obtained from all participants involved in the experiment. All participants were screened by Patient Health Questionnaire^{TM} (PHQ) and only participants with a score lower than 5 were included in the study^{51}.
The affective elicitation was performed by projecting the IAPS images to a PC monitor. The slideshow was comprised of 9 image sessions, alternating neutral sessions and arousal sessions (see Fig. 3). The neutral sessions consist of 6 images having valence range (min = 5.52, max = 7.08), and arousal range (min = 2.42, max = 3.22). The arousal sessions are divided into LowMedium (LM) and MediumHigh (MH) classes, according to the arousal score associated. Such sessions include 20 images eliciting an increasing level of valence (from unpleasant to pleasant). The LM arousal sessions had a valence range (min = 1.95, max = 8.03), and an arousal range (min = 3.08, max = 4.99). The MH arousal sessions had a valence range (min = 1.49, max = 7.77), and an arousal range (min = 5.01, max = 6.99). The overall protocol utilized 110 images. Each image was presented for 10 seconds for a total duration of the experiment of 18 minutes and 20 seconds.
During the visual elicitation, the electrocardiogram (ECG) was acquired by using the ECG100C Electrocardiogram Amplifier from BIOPAC inc., with a sampling rate of 250 Hz. A block diagram of the proposed recognition system is illustrated in Fig. 4. In line with the CMA model, the combination of two levels of arousal and valence brings to the definition of four different emotional states. The stimuli, with high and low arousal and high and low valence, produce changes in the ANS dynamics through both sympathetic and parasympathetic pathways that can be tracked by a multidimensional representation estimated in continuous time by the proposed pointprocess model. The obtained features are then processed for classification by adopting a leaveoneout procedure.
Algorithms
The ECG signal was analyzed offline to extract the RR intervals^{24}, then further processed to correct for erroneous and ectopic beats by a previously developed algorithm^{52}. The presence of nonlinear behaviors in such heartbeat series was tested by using a wellestablished timedomain test based on highorder statistics^{53}. The null hypothesis assumes that the time series are generated by a linear system. We set the number of laps to M = 8, and a total of 500 bootstrap replications for every test. Experimental results are shown in Table 1. The nonlinearity test gave significant results (p < 0.05) on 27 out of 30 subjects. In light of this result, we based our methodology on Nonlinear Autoregressive Integrative (NARI) models. Nonlinearities are intended as quadratic and cubic functions of the past RR intervals according to the WienerVolterra representation^{54,55}. Major improvements of our approach rely on the possibility of performing a regression on the derivative RR series based on an Inverse Gaussian (IG) probability structure^{37,38,39}. The quadratic nonlinearities contribute to the complete emotional assessment through features coming from the instantaneous spectrum and bispectrum^{56,57}. It is worthwhile noticing that our feature estimation is derived from an equivalent n^{th}order inputoutput WienerVolterra model^{54,55}, thus allowing for the potential estimation of the n^{th}order polyspectra of the physiological signal^{60} (see Materials and Methods section for details). Moreover, by representing the RR series with cubic autoregressive functions, it is possible to perform a further instantaneous nonlinear assessment of the complex cardiovascular dynamics and estimate the dominant Lyapunov exponent at each moment in time^{61}. Indices from a representative subject are shown in Fig. 5. Importantly, the NARI model as applied to the considered data provides excellent results in terms of goodnessoffit, and independence test, with KS distances never above 0.056. A comparison analysis was performed between the simple linear and NARI models considering the Sum of the Squared Distances (SSD) of the points outside the confidence interval of the autocorrelation plot (see Table 1). We report that nonlinear pointprocess models resulted in lower SSD on all the considered subjects. Further results reporting the number of points outside the confidence interval of the autocorrelation plot are shown in the Supporting Information.
To summarize, the necessary algorithmic steps for the assessment of instantaneous ANS responses to shorttime emotional stimuli are as follows: a) extract an artifactfree RR interval series from the ECG; b) use the autoregressive coefficients of the quadratic NARI expansion to extract the inputoutput kernels; c) estimate the instantaneous spectral and bispectral features; d) use the autoregressive coefficients of the cubic NARI expansion and fast orthogonal search algorithm to estimate the instantaneous dominant Lyapunov exponent. All the extracted instantaneous features (see Materials and Methods) are used as input of the classification procedure described below. Of note, since no other comparable statistical models have been advocated for a similar application, goodnessoffit and classification performance of the proposed nonlinear approach are compared with its linear counterpart: the basic linear pointprocess model described in^{37}, is also used here for the first time for the proposed classification analysis. Clearly, in order to perform a fair comparison, the order of such a simple linear IGbased pointprocess model was chosen considering an equal number of parameters that need to be estimated for the nonlinear models.
Moreover, here we report on the usability of other simple pointprocess models having the Poisson distribution as interbeat probability. Such a model gave poor performances in terms of both goodnessoffit and resulting in sufficient reliability to solve the proposed classification problems. All of the algorithms were implemented by using Matlab© R2013b endowed with selfmade code and three additional toolboxes for pattern recognition and signal processing, i.e. LIBSVM^{62}, PRTool^{63} and time series analysis toolbox^{64}.
Classification
To perform pattern recognition of the elicited emotional states, a twoclass problem was considered for the arousal, valence and selfreported emotion: LowMedium (LM) and MediumHigh (MH). The arousal classification was linked to the capability of our methodology in distinguishing the LM arousal stimuli from the MH ones, with the neutral sessions associated to the LM arousal class. The overall protocol utilized 110 images. According to the scores associated to each image, for each subject, the dataset was comprised of 64 examples for the LM arousal class and 40 examples for the MH arousal class. Regarding valence, we distinguished the LM from the MH valence regardless of the images belonging to the neutral classes. This choice is justified by the fact that the neutral images can be equally associated to the LM or MH valence classes. According to the experimental protocol timeline, for each subject, the dataset was comprised of 40 examples for the LM valence class and 40 examples for the MH valence class. For the selfreported emotions, we used labels given by the selfassessment manikin (SAM) report. After the visual elicitation, in fact, each subject was asked to fill out a SAM test associating either a positive or a negative emotion to each of the seen images. During this phase, the images were presented in a different randomized order with respect to the previous sequence.
As a further challenge, a more difficult fourclass problem was solved by defining the recognition of four different emotional states associated with specific regions of the arousalvalence plane. Specifically, given a selective combination of the LM and MH levels of the arousal and valence CMA dimensions, the following four emotional states were distinguished: “sadness”, as a simultaneous effect of LM valence and LM arousal; “anger”, as a simultaneous effect of LM valence and MH arousal; “happiness”, as a simultaneous effect of MH valence and MH arousal; “relaxation”, as a simultaneous effect of MH valence and LM arousal (see Fig. 1).
For each of the mentioned classification problems, a leaveoneout procedure^{91} was performed on the N available features using a wellknown Support Vector Machine (SVM)^{66} as pattern recognition algorithm. Specifically, we used a nuSVM (nu = 0.5) having a radial basis kernel function with γ = N^{−1}, as the training vectors. Results gathered from SVM classifiers are summarized in Tables 1 and 2 and reported as recognition accuracy, i.e. the percentage of correct classification among all classes for each subject and for each of the emotion classification cases.
All the estimated HRV indices coming from the NARI models were processed in order to define the whole feature set F_{NARI}. Such a whole feature set is comprised of five linearderived and eleven nonlinearderived features. In this case, linearderived means that such features, namely the mean and standard deviation of the IG distribution (corresponding to the instantaneous point process definitions of mean and standard deviation of the RR intervals^{37}), the power in the low frequency (LF) band, the power in the high frequency (HF) band, and the LF/HF ratio, are estimated having only knowledge of the linear coefficients of the models. Likewise, the nonlinearderived features are estimated from the quadratic and cubic terms of the NARI model. In particular, features coming from the instantaneous bispectral analysis, namely the mean and the standard deviation of the bispectral invariants, mean magnitude, phase entropy, normalized bispectral entropy, normalized bispectral squared entropy, sum of logarithmic bispectral amplitudes, and nonlinear sympathovagal interactions, are derived from the quadratic formulation (see Supporting Information), whereas the dominant Lyapunov exponents analysis was performed engaging cubic nonlinearities (see Materials and Methods).
In order to enhance the personalization of the system as well as improve the recognition accuracy, given the complete set F_{NARI}, we further derived five subsets obtained by applying feature selection. Specifically, we defined the subset F_{1,NARI} having the linearderived features only, the subset F_{2,NARI} having the linearderived along with the nonlinearderived bispectral features, the subset F_{3,NARI} having the linearderived features alongside the dominant Lyapunov exponent, and the subset F_{4,NARI} having the five most informative features as quantified through the Fisher score^{68}. Of note, the sets verify the following relationships: F_{1,NARI} ⊂ F_{2,NARI} ⊂ F_{NARI}; F_{1,NARI} ⊂ F_{3,NARI} ⊂ F_{NARI}; F_{4,NARI} ⊂ F_{NARI}; firstly, we illustrate in Table 1 the classification results by comparing the performances of the reference linear pointprocess model with the NARI model with feature set giving the best recognition accuracy for each subject. The recognition accuracy of the shortterm positivenegative selfreported emotions improves with the use of the nonlinear measures in 25 cases, with >60% of successfully recognized samples for all of the subjects and a maximum of 89.23% for subject 23. In 19 out of the 30 subjects the accuracies are >70% with a total average accuracy of 71.43% Concerning the LM and MH arousal classification, the recognition accuracy of the shortterm emotional data improves in 24 cases, with >69.51% of successfully recognized samples for all of the subjects and a maximum of 95.12% for subject 22. In 29 out of the 30 subjects the accuracies are >70%, 23 of which are above 80%. The total average accuracy is 83.55%. Concerning the LM and MH valence classification, the recognition accuracy of the shortterm emotional data is improved in 23 cases, with >65.38% of successfully recognized samples for all of the subjects and a maximum of 93.83% for subject 23. In 27 out of the 30 subjects the accuracy is >70%, 13 of which are above 80%. The total average accuracy is 79.15%.
Moreover, we show in Table 2 all the results obtained using the complete feature set F_{NARI}. Although such a set might be not optimized, this choice allows for practical affective computing applications, possibly in realtime, retaining all of the information from both linear and nonlinear characteristics of the heartbeat complex dynamics. Alongside the separate arousal, valence, and selfreported emotions, we also show the accuracy in recognizing the four emotional states defined in the arousalvalence plane. In this case, the recognition accuracy is >67.24% of successfully recognized samples for all subjects, with best performance for subject 23 (94.83%). In 27 out of the 30 subjects the accuracy is >70%, 14 of which are above 80%. The total average accuracy is 79.29%.
As a graphical representation of the results, Fig. 6 shows the complementary specificitysensitivity plots for the three emotion classification cases. The complementary specificity is defined as (1specificity), corresponding to the false positive rate. The area of the maximum rectangle that can be drawn within the lowerright unit plane not including any of the points (each representing the result from one subject) can be defined as the “maximum area under the points” (MAUP): the higher the MAUP, the greater the performance. The comparative evaluation of the maximum area under the points reveals that the pointprocess nonlinear (NARI) features give the best results on all the three emotion classification cases.
Other classification algorithms
The performances given by the proposed point process NARI model were also tested using the following six classification algorithms^{67}: Linear Discriminant Classifier (LDC), Quadratic Discriminant Classifier (QDC), KNearest Neighborhood (KNN), MultiLayer Perceptron (MLP), Probabilistic Neural Network (PNN), Vector Distance Classifier (VDC). LDC and QDC are simple discriminant classifiers, i.e., there is no linear or nonlinear mapping of the feature set. Specifically, LDC finds the linear function that better discerns the feature space, whereas QDC finds the quadratic, thus nonlinear, one. MLP and PNN are common neural networkbased algorithms performing a nonlinear mapping of the feature space and are able to find nonlinear functions discriminating the classes. KNN is a clustering algorithm through which the test set is associated to the most frequent class of the K nearest training examples. VDC is a KNN with K = 1.
Results showed that on the arousal classification, considering features coming from the point process NARI model the recognition accuracy of the shortterm emotional data is improved as follows: 21 cases using LDC, 17 cases using QDC, 23 cases using KNN, MLP, PNN, and VDC, among the total 30 cases. On the valence classification, considering features coming from the point process NARI model the recognition accuracy of the shortterm emotional data is improved as follows: 24 cases using LDC, 18 cases using QDC and KNN, 27 cases using MLP, 22 cases using PNN, and 23 cases using VDC, among the total 30 cases. On the selfreported emotion classification, considering features coming from the point process NARI model the recognition accuracy of the shortterm emotional data is improved as follows: 21 cases using LDC, 18 cases using QDC and PNN, 22 cases using KNN, 24 cases using MLP, and 19 cases using VDC, among the total 30 cases.
We also report that SVM algorithms always have the best recognition accuracy.
Discussion
We have presented a novel methodology able to assess in an instantaneous, personalized, and automatic fashion whether the subject is experiencing a positive or a negative emotion. Assessments are performed considering only the cardiovascular dynamics through the RR interval series on shorttime emotional stimuli (<10 seconds). To achieve such results, we defined an adhoc mathematical framework based on pointprocess theory. The pointprocess framework is able to parametrize heartbeat dynamics continuously without using any interpolation methods. Therefore, instantaneous measures of HR and HRV^{37,38} for robust shorttime emotion recognition are made possible by the definition of a physiologicallyplausible probabilistic model. An innovative aspect of the methodological approach is also the use of the derivative RR series^{69,70} to fit the model. This choice allowed us to remarkably improve the tracking of the affectiverelated, nonstationary heartbeat dynamics. The novel fullyautoregressive structure of our model accounts for the pioneering shorttime affective characterization having knowledge related to both linear and nonlinear heartbeat dynamics. In fact, the quadratic and cubic autoregressive nonlinearities instantaneously associated to estimate the most likely heartbeat accounts for the estimation of high order statistics, such as the instantaneous bispectrum and trispectum, as well as significant complexity measures such as the instantaneous Lyapunov exponents. These novel instantaneous features, together with an effective protocol for emotion elicitation and the support of machine learning algorithms, allowed to successfully solve a comprehensive fourclass problem recognizing affective states such as sadness, anger, happiness, and relaxation.
Unlike other paradigms developed in the literature for estimating human emotional states^{19}, our approach is purely parametric and the analyticallyderived indices can be evaluated in a dynamic and instantaneous fashion. The proposed parametric model requires a preliminary calibration phase before it can be effectively used on a new subject. As the methodology is personalized, in fact, the model and classifier parameters have to be estimated using HRV data gathered during standardized affective elicitations whose arousal and valence scores lie on the whole CMA space. Moreover, a fine tuning of some of the model variables, e.g. the sliding time window W and the model linear and nonlinear orders, could be necessary to maximize the performance. Of note, considering shorttime RR interval series gathered from 10 seconds of acquisition, currently used standard signal processing methods would be unable to give reliable and effective results because of resolution or estimation problems. In particular, parametric spectral estimation such as traditional autoregressive models would require interpolation techniques and would not give a goodnessoffit measure, thus making the parameter estimation simply unreliable. In addition, nonparametric methods such as periodogram or Welch spectral estimation require stationarity and, considering the 10 second time window of interest, would give a frequency resolution of only 10^{−1} Hz which is not sufficient to reliably assess spectral powers up to the low frequency spectral components (0.04–0.15 Hz). Fast emotional responses, such as the ones considered in several recent studies^{13,14,97,98}, call for estimation algorithms that can perform a reliable emotional assessment at time resolution as low as 2 seconds. In such cases, our methodology would be able to accurately characterize the emotional response by providing critical high resolution renditions. Accordingly, the methodology proposed here represents a pioneering approach in the current literature and can open new avenues in the field of affective computing, clinical psychology, and neuroeconomics.
Taking advantage of the standard subdivision in LF and HF frequency ranges in the bispectral domain (see Fig. 1 in the supporting information), our method introduces novel nonlinear indices of heartbeat dynamics directly related to higher order interactions between faster (vagal) and slower (sympathovagal) heartbeat variations, thus offering a new perspective into more complex autonomic dynamics. It has been demonstrated, in fact, that the cardiac and respiratory system exhibit complex dynamics that are further influenced by intrinsic feedback mechanisms controlling their interaction^{71}. Cardiorespiratory phase synchronization represents an important phenomenon that responds to changes in physiological states and conditions, e.g. arousal visual elicitation^{48}, indicating that it is strongly influenced by the sympathovagal balance^{72}. We pointed out that lower performance was reported in seven subjects when using nonlinear models. This outcome is neither correlated with goodnessoffit measures nor with the nonlinearity test results on the RR interval series. Possible explanations may span from confounding factors due to the fast protocol pace, to intrinsic individual characteristics of the emotional reaction to stimuli, or simple nonreported distraction from the task by the subject. Further investigations on these factors will be the subject of future studies. Since the proposed pointprocess framework allows the inclusion of physiological covariates such as respiration or blood pressure measures, future work will focus on further multivariate estimates of instantaneous indices such as features from the dynamic crossspectrum, crossbispectrum^{73}, respiratory sinus arrhythmia^{39}, and baroreflex sensitivity^{73} in order to better characterize and understand the human emotional states in shorttime events. We will also direct our efforts in applying the algorithms to a wide range of experimental protocols in order to validate our tools for underlying pathophysiology evaluation, as well as explore new applications on emotion recognition that consider a wider spectrum of emotional states.
The primary impacts are in the affective computing field and in all the applications using emotion recognition systems. Our methodology, in fact, is able to assess the personal cognitive association related to a positive and a negative emotion with very satisfactory results. The emotional model used in this work, i.e., the CMA, accounts for the emotional representation by means of the selective estimation of arousal and valence. Although the recognition accuracy proposed in this work relates to only two levels of arousal, valence and selfreported emotion, oversimplifying the complete characterization of the affective state of a subject, the emotional assessment in shorttime events using cardiovascular information only is a very challenging task never solved before. Using only heartbeat dynamics, we effectively distinguished between the two basic levels of both arousal and valence, thus allowing for the assessment of four basic emotions. An important advantage is that the proposed framework is fully personalized, i.e. it does not require data from a representative population of subjects. These achievements could have highly relevant impacts also in mood disorder psychopathology diagnosis and treatment, since the mood disorder produces an altered emotional response. Hence, monitoring fast emotional response in terms of stimulation time could make a continuous evaluation of disorder progression possible. Presently, emotional state is determined in a clinical setting using questionnaires with limited accuracy and quantitative power. It would be desirable to develop an automated system to determine and to quantify the emotional state. Doing this using a noninvasive physiological and easytomonitor signal such as the electrocardiogram, it would represent an important scientific advancement. From a physiological perspective, the inherent nonlinearties of the cardiovascular systems (e.g. the nonlinear neural signaling on the sinoatrial node^{27}) have been also confirmed by our experimental results. According to the nonlinearity tests^{53}, in fact, 27 out of the 30 RR series resulted to be the outputs of a nonlinear system. Of note, the results from goodnessoffit tests were all positive, demonstrating that the proposed NARI model always performs a good prediction of the nonlinear heartbeat dynamics. We recently demonstrated the crucial role played by nonlinear dynamics in arousal and valence recognition starting from ANS signals^{44}. In agreement with these previous results, here we have introduced instantaneous nonlinear features and improved the accuracy in the majority of the population (23 subjects) over the (also novel) results from the respective linear estimation. This finding is consistent with all the tested classification algorithms. A speculative explanation on the better performances obtained through ANS linear dynamics (observed in 7 subjects) can be related to the functional characteristics of the dynamical nonlinear vagosympathetic interaction, i.e., the socalled bidirectional augmentation^{27}. It is wellknown, in fact, that the ANS control on the cardiovascular dynamics is based on the simultaneous activity and interaction of the sympathetic and vagal nerves. Sunagawa et al.^{27} showed that a perfect balance of these two neural activities leads to a linear control of the mean heart rate with a high potential dynamical gain (variance). On the other hand, when the sympathetic activity prevails on the parasympathetic one and vice versa, the cardiovascular control becomes nonlinear with a reduced dynamical gain. Changes on these timevarying nonlinear vagosympathetic interactions follows a sigmoidal relationship between autonomic nerve activity and heart rate. Accordingly, it could be possible to hypothesize that the vagosympathetic interactions of those 7 subjects showing a more linear ANS dynamics, when elicited with emotional images, are somehow more balanced than in the other subjects.
We also remark on the great improvement, over the random guess of 25%, performed by our methodology in recognizing four different emotional states during shorttime stimuli. Such an achievement surely increases the impact in applicative fields such as affective computing, clinical psycology and neuroeconomics. Moreover, our experimental results support previous studies where instantaneous HRV indices extracted by means of a point process model provided a set of dynamic signatures able to characterize the dynamic range of cardiovascular responses under different physiological conditions^{37}. Therefore, the novel instantaneous nonlinear features could provide better assessment and improved reliability during such physiological responses.
Methods
The experimental protocol for this study was approved by the ethical committee of the University of Pisa and an informed consent was obtained from all participants involved in the experiment.
Nonlinear system identification and nonlinear autoregressive models
A Nonlinear Autoregressive Model (NAR) Model can be expressed, in a general form, as follows: Considering as independent, identically distributed Gaussian random variables, such a model can be can be written as a Taylor expansion: The autoregressive structure of (2) allows for system identification with only exact knowledge on output data and with only a few assumptions on input data. We represent here the nonlinear physiological system by using nonlinear kernels up to the second order, i.e. γ_{0}, γ_{1}(i), and γ_{2}(i, j), and take into account the series of the derivatives in order to improve stationarity^{69,70}. Hence, the general quadratic form of a Nonlinear Autoregressive Integrative (NARI) model becomes: where Δy(k − i) = y(k − i) − y(k − i − 1) and Δy(k − j) = y(k − j) − y(k − j − 1). The quadratic kernel γ_{2}(i, j) is assumed to be symmetric. We also define the extended kernels and as: and link the NARI model to a general inputoutput form, here defined by using the wellknown WienerVolterra^{54} series: where the functions h_{n}(τ_{1}, …, τ_{n}) are the Volterra kernels. Mapping a quadratic NARI model to an nth order inputoutput model^{60} allows, after the inputoutput trasformation of the kernels, the evaluation of all the High Order statistics (HOS) of the system, such as the Dynamic Bispectrum and Trispectrum^{57,74}. In the next paragraph, the definition of the generic variable y(k) is used to represent the firstmoment statistic (mean) of the probabilistic generative mechanism of the heart rate. In other words, the generic formulation of the NARI model parametrizes the autonomic control on the cardiovascular system using linear and nonlinear terms according to the WienerVolterra representation and pointprocess theory.
Pointprocess nonlinear model of the heartbeat
The point process framework primarily defines the probability of having a heartbeat event at each moment in time. Such a probability accounts for the instantaneous estimation of features which are sensible to shorttime emotional variations. Formally, defining t ∈ (0, T], the observation interval, and the times of the events, we can define N(t) = max{k : u_{k} ≤ t} as the sample path of the associated counting process. Its differential, dN(t), denotes a continuoustime indicator function, where dN(t) = 1 when there is an event (the ventricular contraction), or dN(t) = 0 otherwise. The left continuous sample path is defined as . Given the Rwave events detected from the ECG, RR_{j} = u_{j} − u_{j}_{−1} > 0 denotes the j^{th} RR interval. Assuming history dependence, the inverse Gaussian probability distribution of the waiting time t − u_{j} until the next Rwave event is^{37}: with the index of the previous Rwave event before time is the history of events, ξ(t) the vector of the timevaring parameters, the firstmoment statistic (mean) of the distribution, and ξ_{0}(t) > 0 the shape parameter of the inverse Gaussian distribution. Since indicates the probability of having a beat at time t given that a previous beat has occurred at u_{j}, can be interpreted as the expected waiting time until the next event could occur. The use of an inverse Gaussian distribution , characterized at each moment in time, is motivated both physiologically (the integrateandfire initiating the cardiac contraction^{37}) and by goodnessoffit comparisons^{39}. In previous work^{38,39}, the instantaneous mean was expressed as a linear and quadratic combination of present and past RR intervals, based on a nonlinear VolterraWiener expansion^{40}. Here we propose the novel NARI formulation in which the instantaneous RR mean is defined as: In this work, we consider nonlinearities associated to eq. 8 up to the thirdorder. Cubic terms, in fact, allow for the estimation of the dominant Lyapunov exponent, whereas quadratic terms account for the highorder spectral estimation. Considering the derivative RR interval series improves the achievement of stationarity within the sliding time window W (in this work we have chosen W = 70 seconds)^{69}. Since is defined in continuous time, we can obtain an instantaneous RR mean estimate at a very fine timescale (with an arbitrarily small bin size Δ), which requires no interpolation between the arrival times of two beats. Given the proposed parametric model, all linear and nonlinear indices are defined as a timevarying function of the parameters ξ(t) = [ξ_{0}(t), γ_{0}(t), γ_{n}(i_{1}, …, i_{n}, t)] with n = {1, 2, 3}. The unknown timevarying parameter vector ξ(t) is estimated by means of a local maximum likelihood method^{37,75,76}. The model goodnessoffit is based on the KolmogorovSmirnov (KS) test and associated KS statistics (see details in^{37,77}). Autocorrelation plots are considered to test the independence of the modeltransformed intervals^{37}. Optimal order of linear and nonlinear regressions is determined by prefitting the point process model to a subset of the data^{37,75} and comparing scores from KS tests, autocorrelation plots, and the Akaike Information Criterion (AIC) which is defined as follows: where L is the maximized value of the likelihood function for the estimated model, dim(ξ) is the number of parameters in the NARI model. We choose the parameter setup having the autocorrelation plot within the boundaries of 95% confidence interval and minimum AIC value and minimum KS distance. Once the order of linear and nonlinear regressions is determined, the initial NARI coefficients are estimated by the method of least squares^{78}.
Estimation of the inputoutput volterra kernels
The n^{th}order spectral representations are related to the the Volterra series expansion and the Volterra theorem^{54}. In functional analysis, a Volterra series denotes a functional expansion of a dynamic, nonlinear, and timeinvariant function, widely used in nonlinear physiological modeling^{55,79,80}. The quadratic NARI model can be linked to the traditional inputoutput Volterra models by using a specific relationships^{60} between the Fourier transforms of the Volterra kernels of order p, H_{p}(f_{1}, …, f_{n}), and the Fourier transforms of the extended NAR kernels, and . In general, a secondorder NARI model has to be mapped into a infiniteorder inputoutput Volterra model^{60}, although there is the need to truncate the series to a reasonable order for actual application. Here, we choose to model the cardiovascular activity with a cubic inputoutput Volterra by means of the following relationships with the NARI:
Quantitative tools: high order spectral analysis
Our framework allows for three levels of quantitative characterization of heartbeat dynamics: instantaneous timedomain estimation, linear power spectrum estimation, and higher order spectral representation. The timedomain characterization is based on the first and the second order moments of the underlying probability structure. Namely, given the timevarying parameter set ξ(t), the instantaneous estimates of mean RR, RR interval standard deviation, mean heart rate, and heart rate standard deviation can be extracted at each moment in time^{37}. The linear power spectrum estimation reveals the linear mechanisms governing the heartbeat dynamics in the frequency domain. In particular, given the inputoutput Volterra kernels of the NARI model for the instantaneous RR interval mean , we can compute the timevarying parametric (linear) autospectrum^{81} of the derivative series: where . The timevarying parametric autospectrum of the RR intervals is given by multiplying its derivative spectrum by the quantity 2(1 − cos(ω))^{69}. Importantly, previous derivations of the expressions for the autospectrum^{40,73} were possible because the first and secondorder Volterra operators are orthogonal to each other for Gaussian inputs. This property does not hold for orders greater than two^{81}, and in cubic nonlinear inputoutput Volterra systems the autospectrum is estimated by considering also the third order term. By integrating (13) in each frequency band, we can compute the index within the very low frequency (VLF = 0.01–0.05 Hz), low frequency (LF = 0.05–0.15 Hz), and high frequency (HF = 0.15–0.5 Hz) ranges. The higher order spectral representation allows for consideration of statistics beyond the second order, and phase relations between frequency components otherwise suppressed^{57,82}. Higher order spectra (HOS), also known as polyspectra, are spectral representations of higher order statistics, i.e. moments and cumulants of third order and beyond. HOS can detect deviations from linearity, stationarity or Gaussianity. HOS analysis has been proven to be effective in several HRV applications (e.g. arrhythmia recognition^{58,59}). Particular cases of higher order spectra are the thirdorder spectrum (Bispectrum) and the fourthorder spectrum (Trispectrum)^{82}, defined from the Volterra kernel coefficients estimated within the point process framework.
Dynamical bispectrum estimation
Let H_{2}(f_{1}, f_{2}, t) denote the Fourier transform of the secondorder Volterra kernel coefficients. The crossbispectrum (Fourier transform of the thirdorder moment) is^{56,57}: where is the autospectrum of the input (i.e. ). Note that we use the approximation shown in eq. 14 since the equality only strictly holds when the input variables are jointly Gaussian. The analytical solution for the bispectrum of a nonlinear system response with stationary, zeromean Gaussian input is^{83}: Of note, an expression similar to 15 was derived in the appendix of^{84}. The dynamic bispectrum is an important tool for evaluating the instantaneous presence of nonlinearity in time series^{53,56,57}. The bispectrum presents several symmetry properties^{82} and for a real signal it is uniquely defined by its values in the triangular region of computation Ω, 0 ≤ f_{1} ≤ f_{2} ≤ f_{1} + f_{2} ≤ 1^{85}. Within this region, it is possible to estimate several features, which are summarized in Table 1 of the supporting information. The bispectral invariant^{86}, P(a, t) represent the phase of the integrated bispectrum along the radial line with the slope equal to a (with mean and variance σ_{P}_{(a,t)}). The parameter 0 < a ≤ 1 is the slope of the straight line on which the bispectrum is integrated. The variables I_{r}(a, t) and I_{i}(a, t) refer to the real and imaginary part of the integrated bispectrum, respectively. The Mean magnitude M_{mean}(t) and the phase entropy P_{e}(t)^{87} have n = 0, 1, …, N − 1; L as the number of points within the region Ω, Φ the phase angle of the bispectrum, and 1(.) the indicator function which gives a value of 1 when the phase angle Φ is within the range of bin Ψ_{n}. The normalized bispectral entropy (P_{1}(t))^{88} and the normalized bispectral squared entropy (P_{2}(t))^{88} are also considered (between 0 and 1) along with the sum of logarithmic amplitudes of the bispectrum^{89}. As the sympathovagal linear effects on HRV are mainly characterized by the LF and HF spectral powers^{24,90,92,93,94}, we evaluate the nonlinerar sympathovagal interactions by integrating B(f_{1}, f_{2}) in the bidimensional combination of frequency bands. LL(t),LH(t), and HH(t) (see Supporting Information for a graphical representation in the bispectral plane).
Dynamical trispectrum estimation
Brillinger^{95}, Billings^{54}, Priestley^{96}, and others have demonstrated that there is a closed form solution for homogeneous systems with Gaussian inputs. Thus, the transfer function of a morder homogeneous system is estimated by the relation: where the numerator is the m + 1 − n^{th}order crosspolyspectrum between y and x. This result is a generalization of the classical result for the transfer function of a linear system resulting for m = 1. Therefore, the crosstrispectrum (Fourier transform of the thirdorder moment) can be estimated as:
Dominant lyapunov exponent estimation
Cubic nonlinearities, in the framework of the proposed NARI pointprocess model, account for the novel estimation of the instantaneous Lyapunov spectrum. By definition, the generic Lyapunov exponent (LE) of a real valued function f(t) defined for t > 0 is defined as: Let us consider a generic ndimensional linear system in the form y_{i} = Y (t) p_{i}, where Y (t) is a timevarying fundamental solution matrix with Y (0) orthogonal, and {p_{i}} is an orthonormal basis of . The key theoretical tools for determining the IDLE and the whole spectrum of LEs is the continuous QR factorization of Y (t)^{61,65}: where Q(t) is orthogonal and R(t) is upper triangular with positive diagonal elements R_{ii}, 1 ≤ i ≤ n. Then, LEs are formulated as: The cubic NAR model (eq. 8) can be rewritten in an Mdimensional state space canonical representation: By evaluating the Jacobian J(n) over the time series, which corresponds to the matrix Y (t), the LE can be determined using the QR decomposition: This decomposition is unique except in the case of zero diagonal elements. Then the LEs λ_{i} are given by where H is the available number of matrices within the local likelihood window of duration W, and τ the sampling time step. The estimation of the LEs is performed at each time t from the corresponding timevarying vector of parameters, ξ(t)^{61}. Here, the first LE, λ_{1}(t) is considered as the instantaneous dominant Lyapunov exponent (IDLE).
References
 1.
Picard, R. W. Affective computing (MIT Press, Boston, MA, 1997).
 2.
Sauter, D., Eisner, F., Ekman, P. & Scott, S. Crosscultural recognition of basic emotions through nonverbal emotional vocalizations. P. Natl. Acad. Sci. USA 107, 2408–2412 (2010).
 3.
Kessler, R. et al. Lifetime and 12month prevalence of dsmiiir psychiatric disorders in the united states: results from the national comorbidity survey. Arch. gen. psychiat. 51, 8 (1994).
 4.
Wittchen, H. & Jacobi, F. Size and burden of mental disorders in europe: a critical review and appraisal of 27 studies. Eur. neuropsychopharm. 15, 357–356 (2005).
 5.
Tang, Y. et al. Central and autonomic nervous system interaction is altered by shortterm meditation. P. Natl. Acad. Sci. USA 106, 8865 (2009).
 6.
Petrantonakis, P. C. & Hadjileontiadis, L. J. Emotion Recognition From EEG Using Higher Order Crossings. IEEE T. Inf. Technol. B. 14, 186–197 (2010).
 7.
Ko, K. E., Yang, H. C. & Sim, K. B. Emotion recognition using EEG signals with relative power values and Bayesian network. Int. J. Control Autom. 7, 865–870 (2009).
 8.
Li, M., Chai, Q., Kaixiang, T., Wahab, A. & Abut, H. [EEG Emotion Recognition System.]. Vehicle Corpus and Signal Processing for Driver Behavior [Takeda, K. (ed.), Hansen, J. H. L. (ed.), Erdogan, H. (ed.), Abut, H. (ed.)] [125–135] (Springer, New York, 2009).
 9.
Lin, Y. P. et al. EEGBased Emotion Recognition in Music Listening. IEEE T. BioMed Eng. 57, 1798–1806 (2010).
 10.
Petrantonakis, P. C. & Hadjileontiadis, L. J. A novel emotion elicitation index using frontal brain asymmetry for enhanced EEGbased emotion recognition. IEEE T. BioMed Eng. 15, 737–746 (2011).
 11.
Park, K. S. et al. Emotion recognition based on the asymmetric left and right activation. Int. J. Med. Medical Sci. 3, 201–209 (2011).
 12.
Hsieha, S., Hornbergera, M., Pigueta, O. & Hodges, J. R. Brain correlates of musical and facial emotion recognition: Evidence from the dementias. Neuropsychologia 50, 1814–1822 (2012).
 13.
RuizPadial, E., Vila, J. & Thayer, J. The effect of conscious and nonconscious presentation of biologically relevant emotion pictures on emotion modulated startle and phasic heart rate. Int. J. Psychophysiol. 79, 341–346 (2011).
 14.
Tamietto, M. et al. Unseen facial and bodily expressions trigger fast emotional reactions. P. Natl. Acad. Sci. USA 106, 17661–17666 (2009).
 15.
Heller, A. et al. Reduced capacity to sustain positive emotion in major depression reflects diminished maintenance of frontostriatal brain activation. P. Natl. Acad. Sci. USA 106, 22445–22450 (2009).
 16.
Valenza, G. et al. Wearable monitoring for mood recognition in bipolar disorder based on historydependent longterm heart rate variability analysis. IEEE J. Biomed. Health Inform., (2014).
 17.
Valenza, G. et al. Mood recognition in bipolar patients through the PSYCHE platform: preliminary evaluations and perspectives. Artif. Intell. Med. 57, 49–58 (2013).
 18.
Lanata, A., Valenza, G. & Scilingo, E. P. A novel eda glove based on textileintegrated electrodes for affective computing. Med. Biol. Eng. Comput. 50, 1163–1172 (2012).
 19.
Calvo, R. & D'Mello, S. Affect detection: An interdisciplinary review of models, methods, and their applications. IEEE T. Affect. Comp. 1, 18–37 (2010).
 20.
Koolagudi, S. G. & Rao, K. S. Recognition of Emotion from Speech: A Review. Int J Speech Technol 15, 99–117 (2012).
 21.
Agrafioti, F., Hatzinakos, D. & Anderson, A. Ecg pattern analysis for emotion detection. IEEE T. Affect. Comp. 3, 102–115 (2012).
 22.
Wu, C. K., Chung, P. C. & Wang, C. J. Representative Segmentbased Emotion Analysis and Classification with Automatic Respiration Signal Segmentation. IEEE T. Affect. Comp. 3, 482–495 (2012).
 23.
Katsis, C. D., Katertsidis, N. S. & Fotiadis, D. I. An integrated system based on physiological signals for the assessment of affective states in patients with anxiety disorders. Biomed. Signal Proces. 6, 261–268 (2011).
 24.
Task Force of the European Society of Cardiology, the North American Society of Pacing, and Electrophysiology, Heart rate variability: standards of measurement, physiological interpretation and clinical use. Circulation 93, 1043–65 (1996).
 25.
Acharya, U. R., Joseph, K. P., Kannathal, N., Lim, C. & Suri, J. Heart rate variability: a review. Med. Biol. Eng. Comput. 44, 1031–1051 (2006).
 26.
Chua, C. K., Chandran, V., Acharya, U. R. & Lim, C. Application of higher order statistics/spectra in biomedical signalsA review. Med. Eng. Phys. 32, 679–689 (2010).
 27.
Sunagawa, K., Kawada, T. & Nakahara, T. Dynamic nonlinear vagosympathetic interaction in regulating heart rate. Heart Vessels 13, 157–174 (1998).
 28.
Atyabi, F., Livari, M., Kaviani, K. & Tabar, M. Two statistical methods for resolving healthy individuals and those with congestive heart failure based on extended selfsimilarity and a recursive method. J. Biol. Phys. 32, 489–495 (2006).
 29.
Glass, L. Synchronization and rhythmic processes in physiology. Nature 410, 277–284 (2001).
 30.
Costa, M., Goldberger, A. & Peng, C. K. Multiscale entropy analysis of complex physiologic time series. Phys. Rev. Lett. 89, 68102 (2002).
 31.
Poon, C. & Merrill, C. Decrease of cardiac chaos in congestive heart failure. Nature 389, 492–495 (1997).
 32.
Tulppo, M. et al. Physiological background of the loss of fractal heart rate dynamics. Circulation 112, 314 (2005).
 33.
Wu, G. et al. Chaotic signatures of heart rate variability and its power spectrum in health, aging and heart failure. PloS one 4, e4323 (2009).
 34.
Goldberger, A., Rigney, D. & West, B. Science in pictures: Chaos and fractals in human physiology. Sci. Am. 262, 42–49 (1990).
 35.
Sugihara, G., Allan, W., Sobel, D. & Allan, K. Nonlinear control of heart rate variability in human infants. P. Natl. Acad. Sci. USA 93, 2608 (1996).
 36.
Leistedt, S. J. J. et al. Decreased neuroautonomic complexity in men during an acute major depressive episode: analysis of heart rate dynamics. Translational Psych. 1, e27 (2011).
 37.
Barbieri, R., Matten, E., Alabi, A. & Brown, E. A pointprocess model of human heartbeat intervals: new definitions of heart rate and heart rate variability. Am. J. Physiol.Heart C. 288, H424 (2005).
 38.
Barbieri, R. & Brown, E. Analysis of heartbeat dynamics by point process adaptive filtering. IEEE T. BioMed Eng. 53, 4–12 (2006).
 39.
Chen, Z., Brown, E. & Barbieri, R. Assessment of autonomic control and respiratory sinus arrhythmia using point process models of human heart beat dynamics. IEEE T. BioMed Eng. 56, 1791–1802 (2009).
 40.
Chen, Z., Brown, E. & Barbieri, R. Characterizing nonlinear heartbeat dynamics within a point process framework. IEEE T. BioMed Eng. 57, 1335–1347 (2010).
 41.
Valenza, G., Citi, L., Scilingo, E. P. & Barbieri, R. PointProcess Nonlinear Models with Laguerre and Volterra Expansions: Instantaneous Assessment of Heartbeat Dynamics. IEEE T. Signal Proces. 61, 2914–2926 (2013).
 42.
Valenza, G., Citi, L., Scilingo, E. P. & Barbieri, R. Using Laguerre expansion within pointprocess models of heartbeat dynamics: a comparative study. Paper presented at the International IEEE Engineering in Medicine and Biology Conference, San Diego CA, USA, IEEE 2012, August 28–September 1.
 43.
Valenza, G., Citi, L., Scilingo, E. P. & Barbieri, R. Instantaneous Bispectral Characterization of the Autonomic Nervous System through a PointProcess Nonlinear Model. Paper presented at the World Congress on Medical Physics and Biomedical Engineering, Beijing, China, (2012 May 21–31).
 44.
Valenza, G., Lanata, A. & Scilingo, E. P. The role of nonlinear dynamics in affective valence and arousal recognition. IEEE T. Affect. Comp. 3, 237–249 (2012).
 45.
Valenza, G., Allegrini, P., Lanata, A. & Scilingo, E. P. Dominant Lyapunov Exponent and Approximate Entropy in Heart Rate Variability during Emotional Visual Elicitation. Frontiers in Neuroeng. 5, 1–7 (2012).
 46.
Valenza, G., Lanata, A., Scilingo, E. P. & De Rossi, D. Towards a Smart Glove: Arousal Recognition based on Textile Electrodermal Response. Paper presented at the International IEEE Engineering in Medicine and Biology Conference, Buenos Aires, Argentina, IEEE (August 31–September 4).
 47.
Valenza, G., Citi, L., Lanata, A., Scilingo, E. P. & Barbieri, R. A Nonlinear Heartbeat Dynamics Model Approach for Personalized Emotion Recognition. Paper presented at International IEEE Engineering in Medicine and Biology Conference, Osaka, Japan, IEEE (2013, July 3–7).
 48.
Valenza, G., Lanata, A. & Scilingo, E. P. Oscillations of heart rate and respiration synchronize during affective visual stimulation. IEEE T. Inf. Technol. B. 16, 683 (2012).
 49.
Posner, J., Russell, J. & Peterson, B. The circumplex model of affect: An integrative approach to affective neuroscience, cognitive development, and psychopathology. Dev. Psychopathol. 17, 715–734 (2005).
 50.
Lang, P., Bradley, M. & Cuthbert, B. International affective picture system iaps: Digitized photographs, instruction manual and affective ratings. Technical report A6. Univ. Florida, (2005).
 51.
Kroenke, K., Spitzer, R. & Williams, J. The phq9. J. Gen. Intern. Med. 16, 606613 (2001).
 52.
Citi, L., Brown, E. & Barbieri, R. A RealTime Automated Point Process Method for Detection and Correction of Erroneous and Ectopic Heartbeats. IEEE T. BioMed Eng. 59, 2828–2837 (2012).
 53.
Barnett, A. & Wolff, R. A timedomain test for some types of nonlinearity. IEEE T. Signal Proces. 53, 26–33 (2005).
 54.
Billings, S. Identification of nonlinear system  a survey. Proc. IEEE 127, 272–285 (1980).
 55.
Marmarelis, V. Nonlinear dynamic modeling of physiological systems. (WileyIEEE Press, New York, 2004).
 56.
Nikias, C. & Mendel, J. Signal processing with higherorder spectra. IEEE Signal Proc. Mag. 10, 10–37 (1993).
 57.
Nikias, C. & Petropulu, A. Higherorder spectra analysis: A nonlinear signal processing framework. (PTR PrenticeHall, Inc., New Jersey, USA, 1993).
 58.
Chua, K. C., Chandran, V., Acharya, U. R. & Lim, C. M. Cardiac state diagnosis using higher order spectra of heart rate variability. J. Med. Eng. Technol. 32, 145–155 (2008).
 59.
Lanata, A., Valenza, G., Mancuso, C. & Scilingo, E. P. Robust multiple cardiac arrhythmia detection through bispectrum analysis. Expert Syst. Appl. 38, 6798–6804 (2011).
 60.
Le Caillec, J. & Garello, R. Nonlinear system identification using autoregressive quadratic models. Signal Process. 81, 357–379 (2001).
 61.
Citi, L., Valenza, G. & Barbieri, R. Instantaneous estimation of highorder nonlinear heartbeat dynamics by Lyapunov exponents. Paper presented at International IEEE Engineering in Medicine and Biology Conference, San Diego, USA, IEEE (2012, August 28–September 1).
 62.
Chang, C. C. & Lin, C. J. LIBSVM: A library for support vector machines. ACM T. Intell. Syst. Tech. 2, 1–27 (2011). Software available at http://www.csie.ntu.edu.tw/cjlin/libsvm.
 63.
Van der Heijden, F., Duin, R. P. W., De Ridder, D. & Tax, D. M. J. Classification, parameter estimation and state estimation. (Wiley Online Library, New York, USA, 2004).
 64.
Schlogl, A. Time Series Analysis  A toolbox for the use with Matlab. Available online at http://pub.ist.ac.at/%7Eschloegl/matlab/tsa/ (Accessed: 18 February 2014).
 65.
Valenza, G., Citi, L. & Barbieri, R. Instantaneous Nonlinear Assessment of Complex Cardiovascular Dynamics by LaguerreVolterra Point Process Models. Paper presented at International IEEE Engineering in Medicine and Biology Conference, Osaka, Japan, IEEE (July 2013).
 66.
Scholkopf, B., Smola, A., Williamson, R. & Bartlett, P. L. New support vector algorithms. Neural Comput. 12, 1207–124 (2000).
 67.
Duda, R., Hart, P. & Stork, D. G. Pattern classification. (John Wiley & Sons, New York, 2012).
 68.
He, X., Deng, C. & Partha, N. Laplacian score for feature selection. Adv. Neur. In. 18, 507–514 (2005).
 69.
Granger, C. & Joyeux, R. An introduction to longmemory time series models and fractional differencing. J. Time Ser. Anal. 1, 15–29 (1980).
 70.
Chen, Z., Purdon, P., Brown, E. & Barbieri, R. A differential autoregressive modeling approach within a point process framework for nonstationary heartbeat intervals analysis. Paper presented at International IEEE Engineering in Medicine and Biology Conference, Buenos Aires, Argentina, IEEE (2010, August 31–September 4).
 71.
Valenza, G., Lanata, A. & Scilingo, E. P. Improving emotion recognition systems by embedding cardiorespiratory coupling. Physiol. Meas. 34, 449–464 (2013).
 72.
Bartsch, R., Schumann, A., Kantelhardt, J., Penzel, T. & Ivanov, P. Phase transitions in physiologic coupling. P. Natl. Acad. Sci. USA 109, 10181–10186 (2012).
 73.
Chen, Z. et al. Dynamic assessment of baroreflex control of heart rate during induction of propofol anesthesia using a point process method. Ann. Biomed. Eng. 39, 1–17 (2011).
 74.
Mendel, J. Tutorial on higherorder statistics (spectra) in signal processing and system theory: Theoretical results and some applications. Proc. IEEE 79, 278–305 (1991).
 75.
Loader, C. Local regression and likelihood. (Springer Verlag, Berlin, 1999).
 76.
Tibshirani, R. & Hastie, T. Local likelihood estimation. J. Am. Stat. Assoc. 82, 559–567 (1987).
 77.
Brown, E., Barbieri, R., Eden, U. & Frank, L. [Likelihood methods for neural spike train data analysis.]. Computational neuroscience: A comprehensive approach [Feng, J.] [253–286] (CRC Press, Boca Raton, Florida, 2003).
 78.
Westwick, D. & Kearney, R. Explicit leastsquares methods, in Identification of nonlinear physiological systems. (WileyIEEE Press, New York, 2003).
 79.
Akay, M. Nonlinear Biomedical Signal Processing II: Dynamic Analysis and Modeling. (WileyIEEE Press, New York, 2000).
 80.
Jo, J. et al. A nonlinear model of cardiac autonomic control in obstructive sleep apnea syndrome. Ann. Biomed. Eng. 35, 1425–1443 (2007).
 81.
Koukoulas, P. & Kalouptsidis, N. Nonlinear system identification using gaussian inputs. IEEE T. Signal Proces. 43, 1831–1841 (1995).
 82.
Nikias, C. & Raghuveer, M. Bispectrum estimation: A digital signal processing framework. Proc. IEEE 75, 869–891 (1987).
 83.
Nichols, J., Olson, C., Michalowicz, J. & Bucholtz, F. The bispectrum and bicoherence for quadratically nonlinear systems subject to non gaussian inputs. IEEE T. Signal Proces. 57, 3879–3890 (2009).
 84.
Caillec, J. & Garello, R. Asymptotic bias and variance of conventional bispectrum estimates for 2d signals. Multidim. Syst. Sign. P. 16, 49–84 (2005).
 85.
Brillinger, D. & Rosenblatt, M. Asymptotic theory of estimates of kthorder spectra. P. Natl. Acad. Sci. USA 57, 206 (1967).
 86.
Chandran, V. & Elgar, S. Pattern recognition using invariants defined from higher order spectraone dimensional inputs. IEEE T. Signal Proces. 41, 1, 205 (1993).
 87.
Ng, T., Chang, S. & Sun, Q. Blind detection of photomontage using higher order statistics. Paper presented at the International Symposium on Circuits and Systems, Vancouver, Canada, IEEE (2004, May 23–26).
 88.
Chua, K., Chandran, V., Acharya, U. & Lim, C. Application of higher order statistics/spectra in biomedical signals–a review. Med. Eng. Phys. 32, 679–689 (2010).
 89.
Zhou, S., Gan, J. & Sepulveda, F. Classifying mental tasks based on features of higherorder statistics from eeg signals in braincomputer interface. Inform. Sciences 178, 1629–1640 (2008).
 90.
Akselrod, S. [Components of heart rate variability: basic studies.]. Heart rate Variability [Malik, M. (ed.), Camm, A. J. (ed.)] [147–163] (Futura Publishing Company, New York, 1995).
 91.
Kearns, M. & Dana, R. Algorithmic stability and sanitycheck bounds for leaveoneout crossvalidation. Neural Comput. 11, 1427–1453 (1999).
 92.
Malliani, A. Association of heart rate variability components with physiological regulatory mechanisms. Heart rate Variability [Malik, M. (ed.), Camm, A. J. (ed.)] [173–188] (Futura Publishing Company, New York, 1995).
 93.
Malpas, S. Neural influences on cardiovascular variability: possibilities and pitfalls. Am. J. Physiol.Heart C. 282, H6 (2002).
 94.
Ursino, M. & Magosso, E. Role of shortterm cardiovascular regulation in heart period variability: a modeling study. Am. J. Physiol.Heart C. 284, H1479 (2003).
 95.
Brillinger, D. The identification of polynomial systems by means of higher order spectra. J. Sound Vib. 12, 301–313 (1970).
 96.
Priestley, M. Spectral analysis and time series (Academic Press, London, UK, 1981).
 97.
Codispoti, M., Bradley, M. & Lang, P. J. Affective reactions to briefly presented pictures. Psychophysiology 38, 474–478 (2001).
 98.
Smith, J., Low, A., Bradley, M. & Lang, P. J. Rapid picture presentation and affective engagement. Emotion 6, 208–214 (2006).
Acknowledgements
The authors are grateful to Laura Maley for kindly reviewing the paper. The research leading to these results has received partial funding from the Department of Anesthesia, Critical Care & Pain Medicine, Massachusetts General Hospital, and Harvard Medical School, Boston, MA, USA, and European Union Seventh Framework Programme FP7/2007–2013 under grant agreement n 601165 of the project “WEARHAP” and under grant agreement ICT247777 of the project “PSYCHE”.
Author information
Affiliations
Neuroscience Statistics Research Laboratory, Massachusetts General Hospital, Harvard Medical School, Boston, MA, 02114, USA
 Gaetano Valenza
 , Luca Citi
 & Riccardo Barbieri
Department of Brain and Cognitive Sciences, Massachusetts Institute of Technology, Cambridge, MA 02139, USA
 Gaetano Valenza
 , Luca Citi
 & Riccardo Barbieri
Department of Information Engineering and Research Centre E Piaggio, University of Pisa, Pisa, Italy
 Gaetano Valenza
 , Antonio Lanatá
 & Enzo Pasquale Scilingo
School of Computer Science and Electronic Engineering, University of Essex, Colchester, CO43SQ, UK
 Luca Citi
Authors
Search for Gaetano Valenza in:
Search for Luca Citi in:
Search for Antonio Lanatá in:
Search for Enzo Pasquale Scilingo in:
Search for Riccardo Barbieri in:
Contributions
G.V., A.L. and E.S. designed the experiment, acquired and preprocessed the physiological data. G.V., L.C. and R.B. processed and analyzed the dataset using nonlinear pointprocess models. All authors wrote and reviewed the manuscript.
Competing interests
The authors declare no competing financial interests.
Corresponding author
Correspondence to Gaetano Valenza.
Supplementary information
PDF files
 1.
Supplementary Information
Supplementary Information
Rights and permissions
This work is licensed under a Creative Commons AttributionNonCommercialNoDerivs 3.0 Unported License. The images in this article are included in the article's Creative Commons license, unless indicated otherwise in the image credit; if the image is not included under the Creative Commons license, users will need to obtain permission from the license holder in order to reproduce the image. To view a copy of this license, visit http://creativecommons.org/licenses/byncnd/3.0/
To obtain permission to reuse content from this article visit RightsLink.
About this article
Further reading

Genetic algorithm for the optimization of features and neural networks in ECG signals classification
Scientific Reports (2017)

Fusion of heart rate variability and pulse rate variability for emotion recognition using lagged poincare plots
Australasian Physical & Engineering Sciences in Medicine (2017)

Indices from lagged poincare plots of heart rate variability: an efficient nonlinear tool for emotion discrimination
Australasian Physical & Engineering Sciences in Medicine (2017)

Emotional states as mediators between tinnitus loudness and tinnitus distress in daily life: Results from the “TrackYourTinnitus” application
Scientific Reports (2016)

Conversations between self and self as Sigmund Freud—A virtual body ownership paradigm for self counselling
Scientific Reports (2015)
Comments
By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.