Wearable Emotion Classification System Based On Flexible Fabric ECG Wristband

With the development of sensor equipment and material technology,wearable devices that collect ECG signals are increasingly used in the medical field.The collected ECG signal is used as an important input signal in emotion recognition,which can help people understand and adjust their emotional state in time.Therefore,the research of this article is carried out around the design of the daily ECG detection equipment and the recognition of the subject's emotional state.In the field of wearable emotion classification,the collection of ECG signals is simple and less disturbed,so it has become an important data source in daily emotion recognition.The current research has the following problems:(1)The multi-lead ECG sensor is cumbersome to wear,redundant and uncomfortable to wear,which is not conducive to daily monitoring applications;(2)In the current research,there is a lack of the ECG data set established by the wearable device,which has little research on the emotion classification of wearable ECG,and cannot verify the specific effects of this type of equipment in the application of emotion recognition;(3)In the experiment of emotion classification of ECG signals,the effectiveness of manual features is limited and the accuracy is insufficient.In response to the above problems,this article mainly carried out the following work:(1)Designed a flexible fabric ECG wristband based on dry electrodes,and carried out a systematic test experiment to verify its performance in collecting ECG signals.The wristband combined the ECG electrodes,which reduces redundancy and improves comfort.It can achieve a collection effect equivalent to that of wet electrodes.The average correlation is92.30%±7.24%,and it is easy to integrate into daily monitoring equipment;(2)This paper established a positive and negative emotion classification system based on a flexible fabric wristband,carried out ECG data collection experiments under standard emotion-induced materials,obtained effective data from 30 subjects,and established the positive and negative sentiment data set,and based on the data set to extract wavelet,heart rate and peak features,combined with the extreme gradient boosting tree XGBoost reached an average accuracy of85.92%;(3)Aiming at the problem of limited effectiveness of manual features,this paper proposes a one-dimensional convolutional network emotion classification framework based on residual block and channel attention mechanism.The average accuracy of this framework on the self-built emotion data set has reached 89.57%.Compared with the best result of manual features,it is increased by 3.65%,and the effectiveness of the network is proved by the comparison experiment with the classic network.In summary,this paper designs an ECG wristband based on flexible fabrics,forms a wearable emotion classification system based on this wristband,and establishes an ECG emotion data set under standard emotion-induced materials.Machine learning and deep learning extract features for classification,and achieve a better emotion prediction effect,which provides a possibility for the application of ECG wearable devices in the field of emotion recognition.