Anomaly Detection Based on Feature Reconstruction from Subsampled Audio Signals

We aim to reduce the cost of sound monitoring for maintain machinery by reducing the sampling rate, i.e., sub-Nyquist sampling. Monitoring based on sub-Nyquist sampling requires two sub-systems: a sub-system on-site for sampling machinery sounds at a low rate and a sub-system off-site for detecting anomalies from the subsampled signal. This paper proposes a feature reconstruction method for enabling anomaly detection from the subsampled signal. The method applies a long short-term memory-(LSTM)-based network for reconstructing features. The novelty of the proposed network is that it receives the subsampled time-domain signal as input directly and reconstructs the feature vector of the original signal. Experimental results indicate that our method is suitable for anomaly detection from the subsampled signal.