The Research Of Speech Noise Reduction Algorithm And Implementation Based On DSP

During the voyage of the ships,the cabin often subject to external sound of the waves,mechanical noise generated by marine power plant,and other auxiliary device affected caused by vibration and intake and exhaust noise.The environments noise also related to the sailing ships' safety.Excessive noise inside the cabin will seriously affect the normal use of the communication system,thereby affecting the sailing command.Therefore,we need take some measures to reduce noise.According to the number of input sources,speech noise reduction algorithm can be divided into single microphone noise reduction algorithm and microphone array noise reduction algorithm.In this paper,we mainly concentrates on the single microphone algorithm,also implement on the DSP.In this paper,a sub-band based adaptive noise reduction algorithm was proposed to improve speech intelligibility.Weighted overlap-add filter bank and psycho-acoustic model were utilized to analyze and realize the sub-band splitting.The non-linear noise reduction gains were computed with the estimated a posteriori signal to noise ratio(SNR)and a priori SNR.The gain floors were determined with the estimated noise level in dB sound pressure level.The final gains were smoothed between the frames by the peak detector with fine selected attack and release time constants.Experimental results from listening tests showed that about 13% to 35%improvements in intelligibility can be achieved by applying the proposed algorithm to noise corrupted speech.It can meet the voice clarity navigation of the ship in the communication system requirements.However,these algorithms also can be used for other types of noise reduction methods.It also have some practical value on improving the environmental noise and enhancing security of ship navigation.The algorithms of this paper were implemented on the real-time Ezario7100 DSP system which is developed by On Semiconductor and tested by some professional instruments.A quantified gain table is also used to replace the non-linear gain computing when the algorithm was implemented on digital signal processor,and the execution cycle could also be reduced.