Depression recognition base on acoustic speech model of Multi-task emotional stimulus