Research on emotion perception model based on wearable devices

Inhalt:
For the numerical prediction problem in PAD (pleasure; arousal; dominance) dimension emotion prediction and analysis, combined with the characteristics of heart rate variability (HRV), a pad dimension emotion prediction model based on principal component analysis (PCA) and convolutional neural network (Transform) is proposed (PCA-CNN). Firstly, heart rate and heart rate interval data were collected from 20 volunteers in two emotional states, relaxation and anxiety, by means of a multimodal physiological signal acquisition device with music and video induction. The data were then labelled with a PAD scale and the time domain, frequency domain and non-linear characteristics of HRV were extracted by statistical methods such as mean and variance calculations, Welch power spectra and Poincaré scatter plots. The PCA model was then used to reduce the dimensionality of the HRV features. Finally, the reduced HRV features were used as the input features of neural network model for training and prediction. The experimental results show that the PCA-CNN model combined with HRV characteristics has good prediction effect in the three dimensions of PAD, and its average consistency correlation coefficient is 0.51. At the same time, the two prediction methods of CNN based on PCA are compared. The results show that the consistency correlation coefficient of the proposed method is improved by 0.04 and 0.08 respectively compared with the above two methods, indicating that this method can divide emotions carefully, and has a certain complementary role in emotion recognition and analysis combined with wearable devices, it makes it possible to identify and predict emotions in daily life.