Analysis Of Music Modulation On Negative Emotions Based On ECG Signal

As is known to all, in people’s daily life, music not only can inspire people’s corresponding emotion, but also can modulation people’s emotions. However, in the field of study the physiological signal features’ changes during the emotion modulation is fresh. And this research field belongs to Affective Computing, which is a highly integrated emerging research field. The purpose of Affective Computing is to give the computer ability to recognize and understand human emotion, and then build the harmonious interaction environment. In the field of Affective Computing, physiological signals as one of the most important research medium have been shown having the ability to reflect human emotion truly and objectively. Therefore, in this paper, based on the relevant research of Affective Computing, using the ECG signal to analysis the effect on human emotion aroused by music, and the function of regulating negative emotions of music.In this paper, the main research contents and conclusions are as follows:(1) Selected the effective music material, including cheerful, excitement, fear and sadness, four kinds of emotions in all. Designed the music emotion induced experimental scheme, then collected the ECG signals during these four types of music and without music(which collected as for baseline signals) to build up the database of music evoked ECG signal.Design the music emotion induced experimental scheme, collected under the four types of music and the baseline condition by test data of ecg signals, set up the passion of the music evoked ecg signal database.(2) On analyzing the obtained ECG signals, we adopts the method of combining linear and nonlinear characteristics to effectively extract 38 features in total. These features include Wavelet Transform coefficient features(DB3 4 layer wavelet decomposition, extraction the mean, variance and energy characteristics of each layer of high frequency and low frequency coefficient), Embedding Dimension and Time Delay, maximum Lyapunov index and Lempel- Ziv complexity, Approximate Entropy, Hurst index quantitative analysis and Recursive Diagram, etc. Then, in order to classify the different emotions, between music and calm, between music and music, we use Relief feature selection algorithms to weight and sort those 38 features. Then we get the optimal feature combination for the various classification tasks. Finally, in the problem of classifying the four emotions, combining the use of optimal feature combination and the SVM classifier, we obtained that the average recognition rate is 74%, and established the affective classification model to recognize these four emotions. The affective classification model will be used to detect the affective states in real time in the following affective regulation research.(3) In researching the affective regulation problem, the induction-regulation experiment was designed. Using a scary music segment to evoke subjects’ fear firstly, we then use pleasure music segment to regulate their emotions for subjects in the experimental group, and we did not any operations on the subjects in the control group. During the induction and regulation stages, we recorded their ECG signals. Finally, the recorded signals formed the emotion regulation database.(4) Based on the affective classification model and the emotion regulation database, firstly, we determined whether fear were aroused in subjects. And then we input the ECG signal segment of the regulation state, which is determined by a fixed time interval to the affective classification model to see affective states for each subject. Comparing the subjects’ affective states, we found that the time duration for a subject in fear in the experimental group is significantly lower than the time duration of a subject in the control group. In order to further illustrate the regulation function of music segments, we compared each feature in the optimal feature combination between the experimental and control group, and obtain the consistent result. The result shows that the modulation on negative motions compared under cheerful music and natural state, the characteristics of the ECG signal have the same change trend. But the trend in the music state is usually early than the natural state.