Study on Recognizing Anxiety and Depression from ECG Signals Under Six Emotional Evocations

Anxiety and depression are two common negative emotions that can occur independently or coexist with a high probability. When these two emotions coexist, there are significant changes in the activity of the autonomic nervous system compared to when anxiety or depression occurs alone. Ignoring the differences in neural activity between coexisting anxiety and depression versus their individual presence may lead to inaccurate identification results in both anxiety and depression recognition. In this paper, we consider the situation where anxiety and depression coexist and conduct research on identifying anxiety and depression using electrocardiogram (ECG) signals. We propose a Deep Information Extraction and Shared Network (DIESN) architecture that learns deep features from manually extracted characteristics, achieves selective information sharing within deep features, and finally obtains an accuracy of 84.18% for anxiety classification and 88.27% for depression classification through the respective classification structures. This architecture overcomes the limitations of unidimensional recognition of anxiety or depression and improves the overall recognition performance of the model by jointly identifying anxiety and depression, demonstrating sufficient competitiveness in the recognition tasks for both emotions.