作　　者： (黄亮）; (潘平）; (周超）;

机构地区： 贵州大学计算机科学与技术学院,贵阳550025

出　　处： 《计算机应用》 2017年第9期2617-2620,2630,共5页

摘　　要： 针对语音信号的非结构化特点,提出了一种基于量子隧穿效应的说话人真伪鉴别方法。以量子隧穿效应为理论依据,首先,在分析语音信号分帧的量子特性基础上,将每一帧语音信号看作一个量子态,实现算法的量子化;然后,利用势垒能分离能量的特性,通过构建势垒组以提取信号的能量谱特征,并以此作为特征参数;最后,通过高斯混合模型(GMM)进行语音信号建模,完成说话人的真伪鉴别。仿真结果表明,相对于传统方法,利用量子隧穿效应理论实现说话人鉴别可以有效降低算法的复杂度,提高识别的识别率和可靠性,为量子信息理论和说话人真伪鉴别方法提供了新的研究途径。 Aiming at the unstructured characteristics of speech signal, a method of speaker authentication based on quantum tunneling effect was proposed. Based on quantum tunneling effect, the quantum properties of speech signal framing analyzed, and each speech signal frame was regarded as a quantum state, and the quantization of the algorithm was realized. And then the potential barrier was used to separate the energy characteristics. The barrier group was constructed to extract the energy spectrum characteristics of the signal and used it as the characteristic parameter. The speech signal modeling was finally carried out by the Gaussian Mixture Model （GMM） to complete the authentication of the speaker. The simulation results show that compared with the traditional method, the use of quantum tunneling theory to achieve speaker identification can reduce the complexity of algorithm effectively, improve the discrimination and provide a new direction for speaker authentication and quantum information theory.

关 键 词： 说话人 真伪鉴别 量子隧穿效应 高斯混合模型