Research on multimodal depression detection method based on BiGRU and BiLSTM

Depression is one of the most common mental illnesses in the world today. Unlike anxiety in daily life, depression is often accompanied by prolonged low mood, slow thinking, unresponsiveness and difficulty in self-regulation. In severe cases, it can affect life and even lead to death. In this paper, a multimodal depression classification model based on BiGRU and BiLSTM is proposed in the publicly available Chinese dataset EATD-Corpus. Audio and text features are extracted using the vggish model and elmo respectively. The features are not fused. After the audio and text features are trained separately for detection, BiGRU and BiLSTM are adaptively weighted and fused to detect depression. The method has a precision value of 0.66, an F1-score value of 0.77 and a recall value of 0.97. The experimental results show that the performance of the method has been improved.