Development of speech enhancement algorithm for digital hearing aid users in diverse noise environment

Abstract

디지털 보청기 환자들은 잡음 상황에서 언어인지에 불편을 호소한다. 잡음 상황에서의 언어 인지 향상을 위해 다양한 음성 강조 알고리즘들이 디지털 보청기에 적용되어 오고 있다. 본 연구에서는, 개선된 스펙트럼 차분과 컴팬딩 기법을 이용한 디지털 보청기의 음성 강조 알고리즘을 제안하였다. 잔여 잡음을 제거하기 위해 차분 요소에 기반하여 추정된 잡음 스펙트럼의 가중치를 조정하였으며, 포먼트들을 강조하기 위해 음성 존재 여부에 기반하여 포먼트 주파수의 각 채널에 컴팬딩 기법을 적용하였다. 잔여 잡음 현상을 줄이고 약한 음성 성분들을 유지하면서 잡음 제거가 이루어졌다.; Digital hearing aid users often complain of difficulty in understanding speech in the presence of background noise. To improve speech perception in a noisy environment, various speech enhancement algorithms have been applied in digital hearing aids. In this study, a speech enhancement algorithm using modified spectral subtraction and companding is proposed for digital hearing aids. We adjusted the biases of the estimated noise spectrum, based on a subtraction factor, to decrease the residual noise. Companding was applied to the channel of the formant frequency based on the speech presence indicator to enhance the formant. Noise suppression was achieved while retaining weak speech components and avoiding the residual noise phenomena. Objective and subjective evaluation under various environmental conditions confirmed the improvement due to the proposed algorithm. For objective evaluation, we tested segmental SNR and Log Likelihood Ratio (LLR) which have higher correlation with subjective measures. The proposed method has the highest Segmental SNR and the lowest LLR among tested methods. In addition, we confirmed by spectrogram that the proposed method significantly reduced the residual `noise and enhanced the formants. For subjective evaluation, Mean Opinion Score (MOS) that represented the global perception score was tested to 10 normal persons and 9 hearing aid users. This produced the highest quality speech using the proposed method. Also, we measured Speech Reception Threshold (SRT) of hearing aid users based on the 50%-correct point in terms of SNR. The majority of hearing aid users show the result that the proposed method is possible to understand the speech in the lowest SNR environment. Statistical comparisons of the tested methods were performed using the Wilcoxon signed rank test with 0.05 significant levels. We confirmed that every speech enhancement method has a significantly effect and the proposed method has a largest effect in white noise environment, but the proposed method only has a significantly effect in car interior noise environment. In babble noise environment, the spectral subtraction using MCRA doesn’t have an effect and the proposed method has a largest effect. There are significant differences among speech enhancement methods. For the real-time implementation, the proposed algorithm was implemented to the DSP board with commercial TI DSP processor. The performance of the proposed method was evaluated using commercial microphone and receiver of hearing aids, power amplifier, data acquisition module, 2cc coupler and recording microphine in anechoic test box. The processed spectrogram using real-time processing shows the similar pattern with the spectrogram using computer simulation. All of the results using objective, subjective and real-time evaluation show that the proposed speech enhancement algorithm is beneficial for hearing aid users in diverse noisy environments.