Behavioral Recognition of Mice Based on a Deep Network

In recent years, mouse behavior recognition has been widely used in animal behavior experiments, which is great significance to study ethology, research of animal learning and memory and analysis of drug action. However, how to quickly and accurately determine and recording the behavior of mice is still a challenge. Using manual record of mouse behavior is not only time consuming but also low accuracy; the model obtained by training C3D convolutional neural network is used for prediction and the accuracy of the obtained results cannot meet the requirements. In this paper, according to an “open field” experiment scheme in animal behavior experiments, an image preprocessing method that integrates multi-temporal mouse action pictures is proposed. With the theoretical research of deep learning convolutional neural network, the target of mouse behavior classification is finally completed. The preliminary analysis results show that this method has the characteristics of accurate classification, high accuracy and low cost. This method can obtain mouse behavior sequence by continuous batch recognition, which greatly improves the recognition efficiency.