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...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%.展开更多
A new method in digital hearing aids to adaptively localize the speech source in noise and reverberant environment is proposed. Based on the room reverberant model and the multichannel adaptive eigenvalue decompositi...A new method in digital hearing aids to adaptively localize the speech source in noise and reverberant environment is proposed. Based on the room reverberant model and the multichannel adaptive eigenvalue decomposition （MCAED） algorithm, the proposed method can iteratively estimate impulse response coefficients between the speech source and microphones by the adaptive subgradient projection method. Then, it acquires the time delays of microphone pairs, and calculates the source position by the geometric method. Compared with the traditional normal least mean square （NLMS） algorithm, the adaptive subgradient projection method achieves faster and more accurate convergence in a low signal-to-noise ratio （SNR） environment. Simulations for glasses digital hearing aids with four-component square array demonstrate the robust performance of the proposed method.展开更多
文摘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%.
基金Supported by the National Natural Science Foundation of China (60872073)~~
文摘A new method in digital hearing aids to adaptively localize the speech source in noise and reverberant environment is proposed. Based on the room reverberant model and the multichannel adaptive eigenvalue decomposition （MCAED） algorithm, the proposed method can iteratively estimate impulse response coefficients between the speech source and microphones by the adaptive subgradient projection method. Then, it acquires the time delays of microphone pairs, and calculates the source position by the geometric method. Compared with the traditional normal least mean square （NLMS） algorithm, the adaptive subgradient projection method achieves faster and more accurate convergence in a low signal-to-noise ratio （SNR） environment. Simulations for glasses digital hearing aids with four-component square array demonstrate the robust performance of the proposed method.
