Signal recognition of the optical fiber vibration sensor based on two-level feature extraction

Abstract

To deal with the high false alarm rate in optical fiber Michelson interferometer, a two-level feature extraction algorithm based on threshold-crossing rate and sparse auto-encoders is proposed. The threshold-crossing rate algorithm is used as the first level feature extraction to identify whether vibration occurs. If vibrations occur, the sparse auto-encoders algorithm is applied to extract high dimension features of vibration signals, and then the extraction feature will be sent to a classifier to recognize vibration pattern. Experiment results show that this method can effectively identify five kinds of vibrations and reduce false rate.