Attention state detection method based on deep neural network

The invention discloses an attention state detection method based on a deep neural network. The method comprises the following steps: acquiring a video, performing face detection F, and calling an algorithm GF to obtain an attention signal s (t); counting the standard deviation svariation of the s (t) and the mean value vmean of the change speed, and obtaining the state of attention through the flow: when svariation is less than or equal to th1, determining that the state is focused; when svariane is greater than th1 and less than or equal to th2 and vmean is greater than th3 and less than or equal to th4, determining distraction; when the svariation is smaller than or equal to th5 and smaller than or equal to vmean, the impulse is determined; otherwise, determining to be random. Therefore, objective evaluation of different attention states is realized by adopting one camera. 一种基于深度神经网络的注意状态检测方法，包括以下步骤：视频采集、人脸检测F并调用算法G_F得到注意信号s(t)；统计s(t)的标准差svariance和变化速度的均值vmean，通过这样的流程得到注意的状态：当svariance≤th1,断定为专注；当th1＜svari