Modern feature extraction methods and learning algorithms in the field of industrial acoustic signal processing

Abstract

Identification of acoustic events is a challanging field of signal processing. Fast identification algorithms would be applicable for real-time event detection in industrial projects. Event detection is usually done by classifying a specific feature of windows of time series. This paper studies the application of the novel skeleton method for feature extraction. We compare it with traditional feature extraction methods on high frequency sampled vibration data, which was measured by a Gleeble 3800 thermo-mechanical physical simulator. Barkhausen noise and other background noises are hardening the analysis.