| With the development of society,the problem of aging population has become serious,and people are paying more attention to their own health.Hearing impairment and correction are beginning to receive attention.The ability of hearing loss patients to communicate with the outside world is reduced,which seriously affects their quality of life.Hearing aids are the first choice for hearing loss treatment,and a more comfortable wearing experience has become an urgent need for hearing loss patients.There are many key technologies of digital hearing aids.This thesis mainly studies the design and implementation of a non-equal bandwidth multi-channel loudness compensation algorithm based on wide dynamic range compression(WDRC).This thesis starts from the research on the mechanism of auditory perception,and proposes the design idea of WDRC algorithm by summarizing the causes of hearing impairment and the hearing curve of hearing impaired patients.The sound pressure level is selected as the judgment criterion.Before the algorithm is implemented,we use a decibel meter as the measurement standard value to correct the amplitude of the digital signal received by the hearing aid.This thesis proposes an improved multi-channel loudness compensation algorithm based on WDRC.The algorithm uses the corrected sound pressure level and the patient’s hearing curve as the basis for compression.The hearing threshold of patients with hearing loss is higher than that of normal people,gradually increasing as the frequency increases.The pain threshold is lower than that of normal people,and it gradually decreases as the frequency increases.We divide the0～8 k Hz frequency spectrum into 17 channels according to the frequency-sensitive characteristics of the human ear.Human speech is mainly located between250 Hz～3 k Hz,so when we divide this frequency band,the width of the channel is narrower.The gain is calculated independently for each channel,and the compressed 17 components are synthesized and output to obtain loudness compensated speech.We use the piecewise Lagrangian interpolation polynomial fitting method to map the dynamic range of the original speech to the target interval.Compared with the traditional three-stage loudness compensation model,this fitting method is closer to the patient’s hearing curve.The algorithm divides the p atient’s hearing range into three parts.The first part is the low sound pressure level above the hearing threshold.We use third-order interpolation to fit.The second part is the middle sound pressure level segment located around the most suitable area.We use fifth-order interpolation to fit.The third part is the high sound pressure level below the pain threshold.We use second-order interpolation to fit.The improved loudness compensation model can improve voice quality.This thesis also designs a smoothing step effect algorithm for the problem of excessive gain value difference between channels caused by multi-channel loudness compensation algorithm.We use the weighted average method to process the boundary with narrow channel width and small gain difference,and on this basis,we further use the method of least square fitting to process the boundary with wide channel width and large gain difference.Through simulation experiments,we compared the speech waveforms and spectrograms before and after loud ness compensation.The experimental results show that the loudness compensation algorithm proposed in this thesis can perform personalized compression for the hearing characteristic curves of different patients,Algorithms can also improve the intelligibility of speech,which has high practical value. |
