中文摘要：

端点检测是语音识别中一个重要的环节。当信噪比较低时,传统的基于短时能量和短时过零率的端点检测方法不能有效地工作。由于Teager能量算子TEO（Teager Energy Operator）和差分算法可以有效地抑制噪声,因此,提出了一种基于TEO和差分算法的端点检测方法。实验证明,该算法在信噪比较低的情况下,能够随着环境自适应门限,准确地检测出语音信号。通过对两种不同的端点检测算法的比较,证明了基于差分和Teager能量算子的算法的检测正确率较高。

英文摘要：

Endpoint detection is an important part in speech recognition.Traditional detection algorithm based on short-term energy and short-term zero crossing rate can not work effectively when the SNR is small.Teager energy operator（TEO） and differential algorithm can effectively suppress noise,so in this paper,a method for endpoint detection based on TEO and power spectrum difference is proposed.Experiments show that this method is threshold adaptive to the environment and can detect speech endpoint accurately at low SNR.Through the comparison between this method and the traditional endpoint detection algorithm,it is proved that this method has higher correctness rate in detection.