Research And Realization Of Hearing Aid Fitting Formula Based On Artificial Neural Network

There are a lot of people in the world have hearing loss,wearing hearing aid is the main treatment method in addition to drug treatment,wearing hearing aid before hearing aid needs to be fitted,hearing aid fitting work needs to choose the appropriate hearing aid fitting formula to achieve good hearing compensation effect.However,the current hearing aid fitting work for different hearing loss patients need to choose different hearing aid fitting formula to achieve the optimal fitting effect,which brings great inconvenience to the hearing aid fitting work,therefore,this paper proposes a hearing aid fitting formula based on artificial neural network to solve this problem.In this paper,the hearing threshold of each frequency and basic information of patients with hearing loss were determined as the input of neural network and the gain value of each frequency was as the output of neural network by featur e extraction of the key parameters needed for hearing aid fitting.And the structure of the neural network in this paper is determined to be fully connected neural network by referring to the characteristics of various neural networks,After constructing the neural network structure on Tensor Flow platform,a series of training and optimization were carried out on the neural network by modifying the activation function,the number of hidden layer nodes and the learning rate,the average error between the gain calculated by the hearing aid fitting formula proposed in this paper,the gain calculated by the existing hearing aid fitting formula and the expected gain in the collected test data is analyzed and compared.The results show that the average error between the expected small voice gain and the predicted small voice gain by the proposed fitting formula in this paper is1.32 d B,and the average error calculated by the traditional POGOII,NAL-R and NALRP fitting formula is 6.41,3.54 and 6.19 d B,respectively.The average error between the expected big voice gain and the predicted big voice gain by the proposed fitting formula in this paper is 1.49 d B,and the average error calculated by the traditional Fig6,NAL-NL1,NAL-NL2 and DSL[i/o] fitting formula is 3.67,3.61,2.66 and 4.67 d B,respectively.In order to verify the actual fitting effect of this fitting formula,11 hearingimpaired patients were invited to test the fitting effect of hearing aid.It includes objective test and subjective test for hearing impaired patients after they are fitted with hearing aids.The hearing aids and related fitting software are developed and designed by ourselves to carry out hearing aids fitting for hearing impaired patients.Objective test results showed that the hearing threshold and word recognition score of 11 hearingimpaired patients(22 ears)were improved,the average hearing threshold of 22 ears decreased by 32.27 d B on average,and the word recognition score increased by 21.45%on average.Subjective test results showed that compared to the market at present commonly used Fig6,NAL-NL1,DSL[i/o] fitting formula,more hearing-impaired patients like to choose the hearing aid fitting formula proposed in this paper and NALNL2 fitting formula for hearing aid fitting,which is basically consistent with the situation predicted by software.Therefore,the reliability of the hearing aid fitting formula proposed in this paper based on artificial neural network is verified.