摘要
在兼顾降噪性能和功耗的基础上,提出了一种实时多通道数字助听器降噪算法.首先,将输入信号分解为16个子带,计算每个子带的声压级,并基于估计的声压级来计算子带噪声和语音概率;然后,利用直接判决方法计算子带信号的先验信噪比和后验信噪比;最后,计算子带增益函数以实现自适应降噪.将该算法与改进谱减法、自适应维纳滤波法和调制深度法进行了比较.结果表明:与其他3种算法相比,在10 d B白噪声的情况下,本文算法输出的平均信噪比减少约3d B,主观语音质量评估得分最多提高0.90;在4种噪声环境下其平均主观语音质量评估得分提高0.41;所提算法采用子带声压级计算取代信号功率谱估计,节省了快速傅里叶变换的计算量,其时延较其他3种算法至少降低50%.
A real-time noise reduction algorithm for multi-channel digital hearing aids is proposed based on the balance between noise reduction performance and power consumption. First,the input signal is decomposed into 16 subbands and the sound pressure level（ SPL） of each subband is calculated. Based on the estimated SPL,the subband noise and the speech probability are computed.Then,the priori signal noise ratio（ SNR） and the posteriori SNR of the subband signal are calculated by the direct decision method. Finally,the gain function is calculated to adaptively reduce noises.And the proposed algorithm is compared with the improved spectral subtraction,adaptive Wiener filter and the algorithm based on the modulation depth. The experimental results showthat compared with the other three algorithms,the average SNR of the proposed algorithm decreases by about 3 d B and the perceptual evaluation of speech quality（ PESQ） is at most improved by 0. 90 when the SNR of the white noise is 10 d B. In addition,the average output PESQ is improved by 0. 41 in four kinds of noisy environments. In the proposed algorithm,the estimation of the power spectrum is replaced by the calculation of the subband SPL and the fast Fourier transform computation is reduced,inducing at least 50% decrease of the time-delay compared with the other three algorithms.
出处
《东南大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2016年第1期13-17,共5页
Journal of Southeast University：Natural Science Edition
基金
国家自然科学基金资助项目(61273266
61301219
61375028)
江苏省自然科学基金资助项目(BK20130241)
关键词
降噪
多通道助听器
自适应维纳滤波
声压级
noise reduction
multi-channel hearing aid
adaptive Wiener filtering
sound pressure level
