Drug Addiction Detection Algorithm Based on CNN-BN

There is currently few related research in the field of drug detoxification to objectively quantify the degree of drug addiction among drug addicts. This paper designs an experimental paradigm, using a high-density near infrared spectroscopy device to collect forehead near infrared data on 30 drug addicts, and has sorted drug addicts through scientific researchers and hospital experts (10 mild, 10 moderate, and 10 severe).Data were obtained from subjects exposed to pictures of drugs. Aiming at these data sets, a drug addiction detection algorithm based on convolutional neural network and batch normalization (CNN-BN) is designed in this paper to classify mild, moderate and severe addiction levels. we apply the convolutional neural network to extract features in the near-infrared spectroscopy data of drug addicts. In addition, batch normalization layer is added to the convolutional neural network, which greatly improves the training speed and reduces the sensitivity to parameters, and the system's robustness is improved. The final three classification results stabilized at 72%. It provides a theoretical basis for the objective quantitative classification of drug addiction degree and the design of corresponding rehabilitation treatment programs based on the degree of drug addiction.