基于变帧长自适应门限的端点检测方法

传统的双门限端点检测算法的门限一般通过经验值来确定。在实际应用中,这种凭经验得到的判决门限往往不能很好地适应环境变化,尤其是在噪声环境下,经验判决门限不能适应背景噪声的变化。将1种自适应门限、窗长动态变化相结合的双门限端点检测算法用于语音端点检测,实验结果表明该方法在常见的弱噪声环境下效果较好,具有一定的鲁棒性。

An Endpoint Detection Method Based on Dynamic Frame and Self-adaptive Threshold

The threshold of traditional endpoint detection algorithms is usually an experiential value, which hardly works well in variational circumstance and cannot be variable under the noise background especially. The double restricted values algorithm with dynamic frame and self-adaptive threshold is applied to endpoint detection of speech signals. Experimental results show the new method has great performance and robustness in the noisecircumstance.