Research On Loudness Compensation And Speech Quality Evaluation For OTC Hearing Aids

Hearing aids is the most effective intervention and rehabilitation method for people with mild to severe hearing loss.However,the tedious fitting procedures and high prices hinder the popularity of traditional hearing aids.In order to help adults with mild to moderate hearing loss to obtain suitable hearing aids without a complicated fitting process,the U.S.Food and Drug Administration(FDA)proposed to establish a new category of over-the-counter(OTC)hearing aids.Currently,OTC hearing aids are still in the promotion stage,and digital hearing aid technology for OTC hearing aids requires further study.The insight into the characteristics and targeted optimization strategies of OTC hearing aids is provided,and the loudness compensation algorithms and quality evaluation methods for compensated speech have been investigated.To further develop OTC hearing aids,this study designs and implements a speech signal processing solution based on fitting-free OTC hearing aid.The main contents and innovations of this study are as follows.1.A personalized hearing loss classification method has been proposed for fitting-free hearing aid,which classifies the different pattern of hearing loss through a data-driven approach.The tedious fitting procedures and a limited number of hearing experts restrict the popularity of hearing aids.A fitting-free method for classifying hearing loss patterns is beneficial to improve the above situation.Existing classification methods of hearing loss that rely solely on audiograms and do not take into account other factors that may affect hearing loss,resulting in significant deviations from the actual gain required by hearing-impaired patients.A new classifying method for hearing loss patterns is proposed,which provides a set of hearing aid parameters for each hearing loss category to achieve the goal of fitting-free.The proposed method takes the insertion gain of the hearing aid as the classification target.The covariates including patient’s gender,age,wearing history,and tonal language are considered implicitly in the model.Then,an unsupervised clustering analysis method is used to classify hearing loss into six categories and supply a set of hearing aids parameters for each category of hearing loss.Patients can achieve hearing aids without fitting by choosing the most satisfied compensation parameters.The proposed method has been proved to be effective by comparing and analyzing the personalized hearing information.Besides,the proposed method is also compared with traditional audiogram classification method.2.A channel-adaptive loudness compensation method is proposed in this study,which could provide more flexible compensation scheme for OTC hearing aids to satisfy the requirement of various patients.The core function of hearing aids is loudness compensation.The hearing aids parameters obtained from the traditional fitting or self-/free-fitting method are applied to speech through loudness compensation algorithms.Conventional filter bank could not achieve high quality signal after compensation owing to ignored characteristics of human hearing and personalized hearing loss.This study has proposed an innovative channeladaptive loudness compensation algorithm,which uses a Gammachirp filter bank for channel division.The Gammachirp filter bank could simulate the working characteristics of the human cochlear basilar membrane and adjust the number of filter bank channels according to the patient’s hearing situation.This method could meet individual compensation requirements by loudness compensation in each adaptive channel.The new channel-adaptive design method considers not only the information from the audiograms but also the relationship between frequency ranges and speech intelligibility.Experimental results have shown that the proposed multi-channel Gammachirp filter bank achieves better speech signal decomposition and synthesis to obtain good performance with fewer channels.The loudness compensation algorithm based on the Gammachirp filter bank effectively improves sentence intelligibility.The sentence recognition rate of the proposed method is approximately 13% higher than that of the system using Gammatone filter bank.3.A non-intrusive method for objective speech quality assessment is proposed for digital hearing aids.For a long time,the compensation effect of hearing aid is mainly evaluated subjectively,and there are fewer studies of objective evaluation.Furthermore,a pure speech signal is generally required as a reference in the existing objective evaluation methods,which restricts the practicality in a real-world environment.An alternative method of speech quality evaluation is in urgent need to help or replace hearing experts to guide the development of OTC hearing aids.Therefore,this paper presents a non-intrusive speech quality evaluation method for hearing aid,which combines the audiogram and weighted frequency information.The proposed model including an audiogram information extraction network,a frequency information extraction network,and a quality score mapping network is used to evaluate the output speech quality of hearing aids.The input includes the audiograms of hearing loss patients and the Log Mel Filter Bank(LMFB)characteristics of input speech,which provide the information concerning the hearing loss of patients and time-frequency information of the input speech,respectively.The experimental results have shown that it is reasonable to input the audiogram and the weight of the information from two frequency bands simultaneously,which can effectively realize the evaluation on the speech quality of the hearing aid.These works have promoted the signal processing algorithms in OTC hearing aids and laid a foundation for the design of OTC hearing aids products with reliable performance,costeffective and independent intellectual property rights in our country.