基于Gammatone滤波器组的客观语音质量评估

语音质量是评价通信系统的一项重要指标。现有的语音质量感知评估算法采用基于Bark谱的感知模型，其算法复杂度较大，并且对于人耳的频率选择性的模拟存在不足。针对这一问题，本文提出一种新的客观语音质量评估方法，采用更加符合人耳听觉特性的Gammatone滤波器组提取特征参数，计算原始语音与失真语音的平均失真距离，并由主观平均意见分值和归一化平均失真距离之间的映射关系求出客观平均意见分值。实验表明，与感知评估方法相比，本文所提出算法的计算复杂度大大降低，同时保持了客观平均意见分值与主观平均意见分值之间的高相关度。

Objective Speech Quality Evaluation Based on Gammatone Filter Bank

Speech quality is an important indicator for communication system evaluation. The Perceptual Evaluation of Speech Quality(PESQ) is based on perception model of Bark spectrum. The algorithm complexity of PESQ is high and it cannot accurately simulate the auditory characteristic of human ear. In this paper, a novel approach for objective speech quality evaluation is proposed. The Gammatone filter bank that accords with human auditory system is used to extract characteristic parameters. And then the average distortion distance of original voice and distorted voice is obtained. Finally, the map between subjective Mean Opinion Score(MOS) value and normalized average distortion distance is established to get the objective MOS value. Experiments show that the complexity of the proposed algorithm is greatly reduced and the relevancy between the objective MOS and the subjective MOS remain high.