RESUMO
To achieve high-quality voice communication technology without noise interference in flammable, explosive and strong electromagnetic environments, the speech enhancement technology of a fiber-optic external Fabry-Perot interferometric (EFPI) acoustic sensor based on deep learning is studied in this paper. The combination of a complex-valued convolutional neural network and a long short-term memory (CV-CNN-LSTM) model is proposed for speech enhancement in the EFPI acoustic sensing system. Moreover, the 3 × 3 coupler algorithm is used to demodulate voice signals. Then, the short-time Fourier transform (STFT) spectrogram features of voice signals are divided into a training set and a test set. The training set is input into the established CV-CNN-LSTM model for model training, and the test set is input into the trained model for testing. The experimental findings reveal that the proposed CV-CNN-LSTM model demonstrates exceptional speech enhancement performance, boasting an average Perceptual Evaluation of Speech Quality (PESQ) score of 3.148. In comparison to the CV-CNN and CV-LSTM models, this innovative model achieves a remarkable PESQ score improvement of 9.7% and 11.4%, respectively. Furthermore, the average Short-Time Objective Intelligibility (STOI) score witnesses significant enhancements of 4.04 and 2.83 when contrasted with the CV-CNN and CV-LSTM models, respectively.
RESUMO
This paper presents a novel improvement in the optical path structure of a three-wavelength symmetric demodulation method applied to extrinsic Fabry-Perot interferometer (EFPI) fiber optic acoustic sensors. The traditional approach of using couplers to construct the phase difference in the symmetric demodulation method is replaced with a new approach that combines the symmetric demodulation algorithm with wavelength division multiplexing (WDM) technology. This improvement addresses the issue of a suboptimal coupler split ratio and phase difference, which can affect the accuracy and performance of the symmetric demodulation method. In an anechoic chamber test environment, the symmetric demodulation algorithm implemented with the WDM optical path structure achieved a signal-to-noise ratio (SNR) of 75.5 dB (1 kHz), a sensitivity of 1104.9 mV/Pa (1 kHz), and a linear fitting coefficient of 0.9946. In contrast, the symmetric demodulation algorithm implemented with the traditional coupler-based optical path structure achieved an SNR of 65.1 dB (1 kHz), a sensitivity of 891.75 mV/Pa (1 kHz), and a linear fitting coefficient of 0.9905. The test results clearly indicate that the improved optical path structure based on WDM technology outperforms the traditional coupler-based optical path structure in terms of sensitivity, SNR, and linearity.
RESUMO
The identification and separation of sources are the prerequisite of industrial noise control. Industrial machinery usually contains multiple noise sources sharing same-frequency components. There are usually multiple noise sources in mechanical equipment, and there are few effective methods available to separate the spectrum intensity of each sound source. This study tries to solve the problem by the radiation relationship between three-dimensional sound intensity vectors and the power of the sources. When the positions of the probe and the sound source are determined, the sound power of the sound source at each frequency can be solved by the particle swarm optimization algorithm. The solution results at each frequency are combined to obtain the sound power spectrum of each sound source. The proposed method is first verified by a simulation on two point sources. The experiment is carried out on a fault simulation test bed in an ordinary laboratory; we used three three-dimensional sound intensity probes to form a line array and conducted spectrum separation of the nine main noise sources. The sound intensity on the main frequency band of each sound source was close to the result of the near-field measurement of the one-dimensional sound intensity probe. The proposed spectral separation method of the sound power of multiple sound sources provides a new method for accurate noise identification in industrial environments.