DBN神经网络的异常心音检测方法研究OA

In response to the problems of complex models,large amounts of feature data,and difficulty in deploying on mobile devices with limited computing resources in existing methods for detecting abnormal heart sounds,this paper proposes a lightweight abnormal heart sound detection algorithm based on hidden semi Markov models and deep belief networks.Using HSMM for precise segmentation of heart sound signals and extracting feature values from multiple dimensions such as time and amplitude reduces the amount of feature data.The experimental results show that the accuracy of this algorithm reaches 0.886,which is 34 times faster than mainstream algorithms while maintaining similar accuracy.