摘要
通过分析语音特征参数的特点和说话人识别的基本方法,利用DSP的硬件平台,以线性预测倒谱系数为特征参数提取算法以及隐马尔可夫模型为建模算法,实现电子语音锁的系统设计。实验结果表明系统在内部模型数小于10时识别精度高,达到安全保密的要求。
The features of voice-character parameters and fundamental speaker recognition methods are analyzed by using linear prediction cepstrum coefficient and hidden Markov model as modeling algorithms. The experimental results show that the designed system has high identity precision when the interior model amount is less than 10 and can achieve the request of safety-secrecy.
出处
《电声技术》
2006年第7期28-31,共4页
Audio Engineering
