Design and Application of a Laconic Heart Sound Neural Network

Abstract

To design a classification algorithm of heart sounds with low hardware requirements and applicability to mobile terminals, this paper proposes a laconic heart sound neural network (LHSNN). First, we propose three requirements that must be met in the LHSNN design. Then, the specific implementation method of the LHSNN is given as follows: 1) Using a spectrogram as the representation of the heart sound features, the size of the heart sound spectrum is determined according to the principle of lossless information. 2) According to the characteristics of the heart sounds and the design requirements, a neural network is selected and deeply analyzed. 3) Through the optimized method, the network structure satisfies the requirements for running on mobile terminals. Finally, the PhysioNet/CinC Challenge 2016 public heart sound database is used as the experimental object in order to establish a heart sound spectrum library. The experimental results show that the LHSNN can obtain the recognition rate of 96.16% and the modify accuracy of 0.8950, and it can also run on mobile terminals. In addition, the LHSNN has been proven to be adaptable by using the open heart sound dataset of the University of Catania. The research in this paper has positive significance for the classification and recognition of heart sounds in the natural environment.INDEX TERMS Heart sound, laconic neural network, sound spectrum, mobile terminal.

Deep Learning Methods for Heart Sounds Classification: A Systematic Review

“…Conventionally, the 1D heart sound signals are first converted into 2D feature maps that represent the time and frequency characteristics of the heart sound signals and satisfy the unified standards for 2D CNN inputs for heart sounds classification. The feature maps most commonly used for heart sounds classification include MFSC [19,25,32,33], MFCC [26,32], and spectrograms [30,31,34]. Rubin et al [29] proposed a 2D CNN-based approach for the automatic recognition of normal and abnormal PCG signals.…”

“…It should be noted that the utilized datasets were balanced, with each class of heart sound containing 200 recordings. Most deep learning-based methods do not utilize a segmentation algorithm to identify S1, S2, systole, and diastole heart sounds, such as [25][26][27][28]30,31,[33][34][35][36][37][38]44,45,[47][48][49][50][51]54,55]. The methods are nevertheless very efficient for automatic heart sounds classification.…”

Time-Frequency Distributions of Heart Sound Signals: A Comparative Study using Convolutional Neural Networks

“…After the 2016 PhysioNet/Computing in Cardiology (CinC) Challenge [19], using CNN or RNN to conduct heart sound classification became the mainstream approach [20][21][22][23][24][25].…”

Assessment of Dual Tree Complex Wavelet Transform to improve SNR in collaboration with Neuro-Fuzzy System for Heart Sound Identification

“…The signal features are usually extracted either from time domain, wavelet domain, frequency domain, or morphological operations [6][7][8][9]. Many types of ANN and DL techniques have been applied on different datasets with a range of accuracy [10][11][12][13].To facilitate the research to develop an efficient and reliable computer algorithm supporting the diagnosis of heart diseases, international databases have been established containing large PCG data sets of HC and NrHS. PhysioNet Challenge 2016 is an example, which was Massachusetts Institute of Technology (MIT) in the Unites States [14].…”