基于TEO能量谱减法的语音增强技术的研究

RESEARCH ON SPEECH ENHANCEMENT BASED ON TEO ENERGY SPECTRAL SUBTRACTION

谱减法是目前减少噪声干扰、提高语音质量的一种有效方法。为了进一步提高谱减法的去噪性能,提出一种基于TEO(Teager Energy Operator)能量的改进谱减法。该方法利用TEO能量对带噪语音进行语音活动检测,区别出噪声段和语音段,对噪声段和语音段分别进行谱减处理,既保证了语音质量,又尽可能地消除了噪声干扰。在F16战斗机噪声环境中对算法性能进行测试,结果表明,该方法提高了输出信噪比,抑制了音乐噪声,具有良好的语音增强效果。

Abstract Spectral subtraction is an efficient way to reduce noise interference and improve speech quality at present. In order to improve the de-noising performance of spectral subtraction, an improved spectral subtraction based on TEO （Teager energy operator） energy is pro- posed. It uses TEO energy to detect the voice activity in noisy speech in order to distinguish the noise segment and the voice segment and car- ties out spectral subtraction processing on them separately. In this way, the speech quality is guaranteed and the noise is removed as much as possible. The performance of the algorithm is tested in F16 fighter noise environment. Experiments show that this method can improve the out- put SNR and suppress the musical noise, it has a better speech enhancement effect as well.