基于听觉感知模型的多通道语音增强系统

提出以模拟听觉感知模型的非均匀滤波器组为基础的多通道语音增强系统,与基于均匀滤波器组的语音增强系统相比,该方法达到相同频率分辨率所需的通道数较少。采用Itakura-Saito距离对系统进行客观评价,仿真结果表明,该系统增强后的语音比均匀多通道系统增强后的语音具有更好的改善效果。     

Dual symbol optimization‐based partial transmit sequence technique for PAPR reduction in WOLA‐OFDM waveform

Weighted overlap and add‐orthogonal frequency division multiplexing (WOLA‐OFDM) is a new waveform proposed recently for meeting the requirements of fifth generation (5G) telecommunication standards. In spite of being a serious 5G waveform candidate, WOLA‐OFDM is exposed to the problem of high peak to average power ratio (PAPR) similar to the other waveforms in which multicarrier transmission strategy is employed. Due to the overlapping nature of WOLA‐OFDM waveform, where the extension of the current symbol is overlapped with the extension of the previous symbol, it will not be efficient to apply conventional PTS (C‐PTS) directly to the WOLA‐OFDM waveform. Therefore, in this paper, we propose dual symbol optimization‐based partial transmit sequence (DSO‐PTS) technique for PAPR reduction in WOLA‐OFDM waveform. In our proposed technique, two adjacent symbols are jointly considered when searching for the optimal data block with minimum PAPR unlike the C‐PTS where the adjacent symbols are optimized individually. In the simulations, our proposed DSO‐PTS technique, C‐PTS, and GreenOFDM that is developed recently by modifying the conventional selective mapping (SLM) method are compared with each other with regard to PAPR reduction performance for different search numbers (SNs). In addition, the effects of DSO‐PTS, C‐PTS, and GreenOFDM on the amount of out of band (OOB) radiation in the power spectral density (PSD) graph of WOLA‐OFDM employing solid state power amplifier (SSPA) is measured for different SNs and input back off (IBO) values. According to the simulation results, our proposed DSO‐PTS technique clearly demonstrates a superior PAPR reduction and PSD performance.     

适用于助听器语音信号处理的DFT调制滤波器组设计

提出了一种近似重构、过采样DFT调制滤波器组的设计方法。根据滤波器组的系统失真和输出混迭方面的要求,构造一个简单的目标函数,并采用等纹波法优化设计原型滤波器。与现有方法相比,该设计方法将复杂的非线性优化问题转化为对单一参数的选取,简化了设计过程,并且可在满足DFT调制滤波器组设计准则的条件下使得通带范围内误差逼近最小。利用加权叠接相加（Weighted Overlap and Add,WOLA）的快速实现结构,使整个系统更加高效、易行,可用于助听器语音信号分频带处理与综合。     

WOLA滤波器组在宽带数字中频处理中的应用

信道化技术是一种宽带数字中频系统中常用于信号处理与分析的技术。多相DFT滤波器组是实现信道化的一种高效结构,但其参数设计的灵活性受到了信道数目与抽取倍数必须相等的限制。利用WOLA滤波器组实现宽带数字中频的信道化设计,可实现与多相DFT滤波器组相同的功能,同时解除了信道数目与抽取倍数的约束关系。仿真结果验证了WOLA滤波器组实现宽带信号信道化的有效性。     

一种模拟听觉感知模型的非均匀滤波器组

结合均匀滤波器组的加权叠接相加法和短时全通变换,设计了一种可以模拟听觉感知模型的非均匀滤波器组。该滤波器组可用于实时信号处理。仿真结果显示,设计的非均匀滤波器组可以模拟Bark频率尺度,且具有比较理想的信号重建功能。