摘要
在虚拟仪器平台上,通过微型麦克与计算机声卡对语音信号进行实时采集,并进行消噪处理和端点检测,以美尔频率倒谱系数及其差分作为特征参数提取方法,应用改进的矢量量化-隐马尔可夫识别算法,设计了与文本无关的说话人识别系统。系统运行结果表明,本文实现的系统运算速度快,且具有较高的识别率。
A microphone and a computer sound card are employed to acquire real-time speech signals and perform noise suppression and endpoint detection on a LABVIEW platform. By using the Mel Frequency Cepstrum Coefficient, speaker recognition＇ s characteristic pmmneter is extracted. The text-independent speaker identification system is designed through Improved VQ （Vector Quantization）-HMM （Hidden Markov Model） algorithm. The results show that it has high computation speed and higher recognition rate.
出处
《山东科学》
CAS
2008年第4期57-61,共5页
Shandong Science
