Implementation of a deep learning model for emotion evaluation based on LSTM psychological and physiological data

Abstract

Evaluating mental status is an important issue that diagnosing depression. Hamilton Depression Rating Scale (HAM-D) is a common method to diagnosis depression. Generally, psychiatrists collect Monthly Mood Chart (MMC) to infer mental status of patients during treatments. However, the processes waste a lot of time. Therefore, our target is to find a method that can evaluate mental status faster. We’d used the constructed platform[15] to collect physiological and psychological data. We’d collected 91 data including 42 remission data and 49 non-remission data. We’d used Electroencephalography(EEG) to train LSTM model, and then got 70% accuracy. This model can automatically infer mood status that helping psychiatrists evaluating. This system had coordinated with two hospitals to refer mood status in the future.