Old people's abnormal behavior detection method based on deep learning

The invention proposes an old people's abnormal behavior detection method based on deep learning, which belongs to the deep learning field. According to the invention, a plurality of sensors are used to acquire the body characteristic information, the position information and the image information of an old person so as to determine the abnormal behaviors of the old person in a jointed detection way, reducing the mis-judgment probability. The method comprises: first, based on the data of the plurality of sensors, preprocessing the signal; inputting the processed data to a well-trained BP neural network to obtain the health condition of the old person; then, according to the original image, preprocessing the image and transmitting the image to a 3D convolution neural network to extract characteristic vectors; and through the Softmax classifier recognizing the multiple behaviors by the old person; and in combination of the position information of the old person as well as his or her duration there and according