Depression Identification from Gait Spectrum Features Based on Hilbert-Huang Transform

Abstract

Depression is a common mental illness, which is extremely harmful to individuals and society. Timely and effective diagnosis is very important for patients’ treatments and depression preventions. In this paper, we treat the trajectory of gait as signal, proposing a new direction to detect the depression with gait frequency features based on Hilbert-Huang transform (HHT). Two groups of participants are recruited in this experiment, including 47 healthy people and 54 depressed patients, respectively. We process the gait data with HHT and build the classification models which verification method is leave-one-out. The best result of our work is 91.09% when the model I is adopted and the classifier is SVM. The corresponding specificity and sensitivity are 87.23% and 94.44% respectively. It verifies that the gait frequency of patients with depression is significantly different from that of healthy people, and the frequency domain features of gait are helpful for the diagnosis of depression.