【摘要】 本文提出了一种新的语音激活检测算法,这种方法基于竞争神经网络,主要应用了自组织特征映射网络并结合学习向量量化算法进行实现,并与其它神经网络算法进行了比较。该算法在多种噪声背景下具有较强的鲁棒性,仿真结果表明,这种基于竞争神经网络的算法优于ITU—T G．729B建议的算法。更多还原

【Abstract】 A new Voice Activity Detection (VAD) algorithm is proposed in this paper. This algorithm is based on the competitive neural networks, especially the Self-organized Feature Mapping (SOM) and the Learning Vector Quantization ( LVQ) network. This algorithm is compared with other neural network algorithms. This algorithm is tested in various noise backgrounds and it is proved to be a robust algorithm under noise environments. The simulation results show that this VAD algorithm which based on the competitive networks has a significant improvement over traditional methods such as ITU - T G. 729B VAD.更多还原