基于竞争网络的语音激活检测研究

A New Voice Activity Detection Algorithm Based On Competitive Networks

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。

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.