Design framework of hybrid ensemble identification network and its application in heart sound analysis

Mixed heart sounds include heart sounds in a state of resting and motion. The analysis of heart sound signals in a state of motion is a difficult problem. (1) First, the mixed heart sound signal was collected by using the shoulder-strap-type heart sound acquisition device designed and made by our research group. The acquisition scheme and data preprocessing method were given, and the characteristics of heart sound signals in a state of motion were analyzed. (2) The design framework of the Hybrid Ensemble Identification Network (HEINet) is proposed, and the design requirements, architecture principles, and detailed design steps are discussed. The design process is simple, fast, and convenient. (3) In this paper, according to the design framework of HEINet, HEINet of the mixed heart sound signal is designed, and the recognition rate of the mixed heart sound signal in biometric authentication has reached 99.1%. Based on this design framework, HEINet of the heart sound signal for the Heart Sounds Catania 2011 heart sound database and HEINet of the electrocardiogram signal for Massachusetts Institute of Technology-Beth Israel Hospital arrhythmia database were designed, and the recognition rates both met the expected requirements. It shows that the design framework of HEINet has obvious universality.