数字助听器中基于两步降噪的多通道频响补偿算法

Multi-channel frequency response compensation algorithm based on two-step noise reduction in digital hearing aids

目的针对数字助听器中现有频响补偿算法在噪声环境下无法满足耳聋者言语可懂度和舒适度的问题,本文提出了基于两步降噪的多通道频响补偿算法。方法该算法首先通过计算通用单通道语音增强的先验信噪比和系统增益,并利用前后语音帧间的相关性,结合自适应先验信噪比平滑因子对先验信噪比估计进行优化,实现两步降噪;然后采用gammatone滤波器对增强后的语音进行非均匀多子带分解,在每个子带中采用两次插值;最后依据耳聋者的听力曲线对每个频带进行频响补偿。结果实验表明,该算法能够有效地去除了噪声对频响补偿的影响,保护了共振峰结构,获得更多增益信息,补偿了患者缺失的语音能量,满足了听力损失者的听力需求。结论基于两步降噪的多通道频响补偿算法有效提高了听障患者感知语音的可懂度和舒适度。

Objective Aiming at the problem that the existing fiequency response compensation algorithm for digital hearing aids in noisy environments can not meet the requirement of the deaf for intelligibility and comfort of digital hearing aids, we proposed the multi-channel frequency response compensation algorithm based on two-step noise reduction in this paper. Methods First, we calculate the prior signal-to-noise ratio and system gain of the common single-channel speech enhancement, and use the correlation between the previous and the subsequent speech fiame to optimize the priori SNR estimation with the adaptive priori SNR smoothing factor so that two-step noise reduction is implemented. Then, the gammatone filter is used to perform nonuniform multi-subband decomposition on the enhanced speech, and interpolation is performed twice in each sub-band. Finally, the fiequency response of each fiequency band is compensated according to the hearing loss curve of the deaf. Results Experiments show that this algorithm can effectively remove the impact of noise on the fiequency response compensation, protect the structure of the fore, ant, obtain more information, compensate the missing speech energy of the patient, and meet the hearing requirement of people with hearing loss. Conclusions The proposed method can effectively improve the intelligibility- and comfort of auditory perception of hearing impaired patients.