Enhanced heart sound anomaly detection via WCOS: a semi-supervised framework integrating wavelet, autoencoder and SVM

Frontiers in Neuroinformatics, Journal Year: 2025, Volume and Issue: 19

Published: Jan. 29, 2025

Anomaly detection is a typical binary classification problem under the condition of unbalanced samples, which has been widely used in various fields data mining. For example, it can help detect heart murmurs when structurally abnormal, to tell if newborn congenital disease. Due low time and high efficiency, most work focuses on semi- supervised anomaly method. However, effect this method not because massive with uneven samples different noise. To improve accuracy sample conditions, we propose new semi-supervised (WCOS) based clustering, combines wavelet reconstruction, convolutional autoencoder, one support vector machine. In way, only distinguish small proportion abnormal sounds huge scale but also filter noise through reduction network, thus significantly improving accuracy. addition, evaluated our using real datasets. When sigma = 0.5, AUC standard deviation WR-CAE-OCSVM 19.2, 54.1, 29.8% lower than that WR-OCSVM, CAE-OCSVM OCSVM, respectively. The results confirmed higher WCOS compared other state-of-the-art methods.

Language: Английский

An unsupervised anomalous sound detection method based on similarity-driven automatic feature selection

Digital Signal Processing, Journal Year: 2025, Volume and Issue: unknown, P. 105123 - 105123

Published: March 1, 2025

Language: Английский