Research On Micro-vibration Detection System Based On Multi-sensor Data Fusion

Laser micro-vibration measurement has the advantages of high precision,remote and non-contact measurement,which is widely used in national defense security,crime fighting,industrial detection,biomedicine and other fields.Laser vibration measurement mainly includes laser reflection method,laser interference method and laser speckle method.Laser reflection method requires the separation of launcher and receiver,and the system is difficult to arrange.Laser interference method requires high coherence of light source and it has a complicated optical path.By contrast,the laser speckle method overcomes the shortcomings of the two methods above.It is an advanced method for micro-vibration detecting because of its simple optical path and easy arrangement.Data fusion is a hot research topic in the field of signal processing by combining multiple data to improve the distinguish ability of the system.We study the application of multi-sensor data fusion in laser speckle micro-vibration measurement,which is innovative and possesses a value of application.Firstly,a variety of Chinese and English voices are used as the micro-vibration excitation source,and a 532nm single-frequency laser is used to irradiate the vibration surface to generate spatial speckle.Under different detection distances,linear CCD is used to record speckle information by multi-channel.Secondly,the collected information is pre-processed and restored to speech signal of degraded.The direct multi-channel fusion,single-channel processing before multi-channel fusion and multi-channel fusion before single-channel processing were used to deal with the degraded speech to explore the best way to improve the demodulation signal performance.Finally,the signal quality of Chinese environment was evaluated by segmental SNR(segSNR),and the signal quality of English environment was evaluated by Perceptual evaluation of speech quality(PESQ).The results show that the three processing methods can improve the signal quality and have good robustness under different test distances.We evaluate the three processing methods with segmental SNR.For none-weight delay-and-sum beamformer(nw_ds)and phase transform delay-and-sum beamformer(phat_ds),multi-channel fusion before single-channel processing is superior to single-channel processing before multi-channel fusion,and single-channel processing before multi-channel fusion is better than direct multi-channel fusion.For phase-error based filtering(pbf),the best method is single-channel processing before multi-channel fusion,multi-channel fusion before single-channel processing is in the middle,the poorest is direct multi-channel fusion.When evaluating the three processing methods with PESQ,multi-channel fusion before single-channel processing is superior to direct multi-channel fusion.And direct multi-channel fusion is better than single-channel processing before multi-channel fusion.