基于隐马尔可夫模型的ATM机用户异常行为识别

摘要

提出一种基于隐马尔可夫模型(hidden markov model,HMM)的ATM机异常行为识别方法.对ATM机前用户存(取)款行为的视频序列用Hu变换提取运动目标的行为特征,采用Baum-Welch算法对正常行为训练并建立隐马尔可夫模型,通过模型输出测试样本序列的概率来识别异常行为.用Matlab对ATM机用户运动行为的模拟视频进行实验仿真,结果表明:该方法对ATM机前的用户行为具有较高的识别率.%A recognition approach of ATM abnormal behavior was presented based on hidden Markov model(HMM). The behavior characteristics of moving object was extracted with Hu transformation from the video train of the deposit (withdrawal) behavior of the user in front of ATM. The Baum-Welch algorithm was used for normal behavior training and the hidden Markov model was set up. Finally, the ATM user's abnormal behavior was recognized by using the probability of test sample sequence of the model output. The simulation video of ATM user's behavior was simulated with Matlab software. The result showed that this method would exhibit a higher recognition rate of ATM user's behavior.