| Hearing is one of the most important perceptions of human beings.The large number of hearing impaired people makes the market more demanding and more comfortable for hearing aids.The loudness compensation algorithm is the basis of many algorithms for digital hearing aids and is the most important of all hearing aid algorithms.In order to better solve the urgent needs of hearing patients to develop a more comfortable and comfortable digital hearing aid,this paper specifically researches and implements a customized loudness compensation algorithm in different frequency channels.In this thesis,four endpoint detection algorithms based on logarithmic spectral distance,cepstrum distance,MFCC parameter cepstrum distance and spectral entropy method are studied and simulated,and then improve the endpoint detection by spectral entropy method to improve the endpoint detection algorithm.Accuracy and stability.Adding the improved spectral entropy endpoint detection to the loudness compensation algorithm,so that the initial noisy signal can be processed to distinguish the starting and ending points of the speech frame from the noisy speech.At this time,the loudness compensation algorithm is used.Real-time compensation can be performed only for the speech frame,which not only improves the accuracy of the loudness compensation algorithm but also reduces the system resource waste generated when the system processes the noise signal.This dissertation improves the design of the traditional single-channel or equal-width multi-channel loudness compensation algorithm.Since the human ear's perception of sound changes with frequency,the frequency band perceived by the human ear is channel-divided.In this paper,the frequency band of 0 Hz~8 kHz is divided according to the division method of human ear perception characteristics,and the 1/3 octave channel division method is used,and the most sensitive 1 kHz~3 kHz is segmented.The frequency bands from 0 Hz ~ 8 kHz are integrated into eight frequency channels of different widths,and a non-equal width channel division method based on human ear perception characteristics is realized.In this thesis,the correction of the sound pressure level and the actual sound pressure level caused by the hearing aid in the sound collection process due to the hardware platform is corrected.The sound pressure level is obtained by the decibel meter and calculation formula for the six different speeches,and then the sound pressure level calculation formula is corrected according to the difference between the two results,and finally the sound pressure level calculation formula based on the hardware platform is obtained.Make data processing more accurate.In this paper,the specific gain is compensated in each channel separately.Newton interpolation method is used to interpolate different data in each channel.The three-stage compensation scheme in traditional loudness compensation is improved,so that the compensation effect is closer.The loss of hearing of the patient.Finally,through simulation,the loudness compensation scheme is obtained,which has high practical value in improving the hearing ability of hearing patients. |
PDF Full Text Request