Research On Shaping Filter Algorithm For Digital Hearing Aids

The research on algorithms for digital hearing aids has great significance in our ownR&D and the production. A filtering algorithm has been proposed in this paper based onthe studying of hearing aids research, clinical audiology and speech signal processingtechnology abroad. Looking forward to achieving the compensation for hearing loss andproviding a new idea for the existing algorithms.According to the change of inputing sound pressure level, setting a plurality ofcompensation curve to fit the nonlinear response of the human ear at different soundpressure levels. So a comprehensive algorithm is designed which can achieve noisecancellation and change the compensation curve adaptively. Firstly, in the manner ofmulti-channel achieve the multi-response curves at different sound pressure levels byadjusting the gain value of each channel. Each channel complies by FIR band pass filter toensure the linear phase. Frequency range of all channels are divided according to thefrequency distribution of the cochlear sensorineural structures, obtaining the best hearingeffect. Secondly, the spectral subtraction noise reduction algorithm eliminate thebroadband noise which does not require in patients’life, improving SNR and making voicemore clearly.During the design of Algorithm,comparing a variety of endpoint detectionalgorithm and using the improved spectral subtraction algorithm, in order to reduce thevoice distortion by noise reduction. Finally, the sound level evaluate algorithm which isdesigned by the principles of Sound Level Meter makes program more flexible and makesmulti-channel compensation curve switch according to the environment. In addition, thecommonly digital signal processing algorithms are used, such as FFT,framing,window-adding, convolution,etc. Filtering algorithm is achieved by the way of fastconvoluteion,in order to improve the efficiency and ensure the real-time performance.Eachalgorithm is designed and implemented by MATLAB,Furthermore, the aid of FDATool inMATLAB design the bandpass filter of the multi-channel filtering algorithm section andsimulation in Simulink. By the way of the frequency analysis, and direct audio player todetermine whether the bandpass filter parameter design is reasonable, high efficiency parameter adjustments and improvements.Considering the algorithm needs to be transplanted to embedded platforms eventually,using TI DSP model produced TMS320C5509A,And building a embedded test platform ofvoice processing combining with TLC320AIC23B audio processing chip.The platformadopts CCS development environment and MATLAB for joint design, not only speeds upthe debug cycle of transplant process but also reduces the difficulty of design. The use ofembedded operating system BIOS which the DSP supports as the carrier of thealgorithm,Can reasonablely dispatch each algorithm by the way of multitasking mode,multiple tasks together with the completion of the processing of speech. The build ofplatform provides better conditions for the study of further improvements and newalgorithms.