摘要
研究一种适合医疗仪器的语音识别算法,采用带噪声端点检测算法、美尔频标倒谱系数(MFCC)特征参数提取算法和整体路径约束DTW算法即ADTW算法,能有效地提高识别率和稳健性。在此为噪声环境下的语音识别提供了理论分析与仿真实验数据。实验结果表明:这种语音识别算法不仅有很高的识别率,而且能减小或者消除噪声所带来的训练模型和测试语音之间的失配。它完全满足医疗仪器对语音识别率的要求。
A speech identification algorithm suit for medical devices is researched,which improves the identification rate and steadiness efficiently by using noisy endpoint detecting algorithm,MFCC character parameter extraction algorithm and the whole path constraint DTW algorithm（namely ADTW mode match algorithm）.The theory analysis and simulation experiment data of speech identification under noise environment are also provided.Simulation experiment results show that this speech identification algorithm has high identification rate,and can decrease or eliminate the mismatch between training pattern and speech test resulting from noise.It can reach all demands of speech identification rate for medical devices.
出处
《现代电子技术》
2010年第10期40-42,48,共4页
Modern Electronics Technique
