Research On Speech Noise Reduction And Extraction Technology Based On Data Fusion And Image Pyramid Processing

Non-contact micro-vibration measurement has the advantages of high measurement accuracy,fast measurement speed,and does not change the state of the measured object.It has unique advantages and application prospects in the fields of aerospace,biomedicine,and industrial agriculture.Among them,the micro-vibration measurement technology based on laser speckle has attracted the attention of relevant researchers due to its advantages of wide spatial coverage,high light wave energy and high sensitivity.The combination of laser speckle and high-speed image acquisition makes this micro-vibration measurement technology simple,fast and rapid.Further,the extraction and amplification of motion characteristics in high-speed video under ambient light illumination has the advantages of strong concealment and high sensitivity,which has aroused great interest from relevant personnel.This paper first reviews the current research status at home and abroad based on laser speckle measurement,and then uses linear CCD as the imaging medium to study the effect of multi-channel data fusion on signal-to-noise ratio and intelligibility.Then,the related application of the steerable complex pyramid method in video image processing is reviewed,the effect of filter segmentation on the image is simulated,and the effectiveness of this method in micro motion extraction is verified experimentally.Specific research contents include:(1)Using linear CCD as the recording medium,a micro-vibration measurement experiment based on laser speckle was carried out.A variety of data fusion and speech evaluation algorithms were used to study the effect of detection distance and sound source type on the fusion result.The speech signal restored by the experiment has a maximum intelligibility of 0.73 and a maximum signal-to-noise ratio of 6 dB.The results show that the data fusion method improves the signal-to-noise ratio more obviously at long measurement distances;the improvement of the signal-to-noise ratio has basically no effect on the change of intelligibility.(2)Using the image decomposition method of the steerable complex pyramid,a micro-vibration measurement experiment was carried out based on its phase information.Under ambient light illumination,a common speaker was used as the excitation source,and office plants and potato chip bags were used as the measured objects.Successful extraction of standard scale vibration signals.Using the higher light reflection characteristics of potato chip bags,the measurement and extraction of singlesource speech,multi-source speech and music speech signals have been successfully achieved.Affected by the brightness of the ambient light,only vibration signals up to 400 Hz can be extracted in the experiment.