论文部分内容阅读
在关键词检出系统中由于替换错误而导致的识别性能下降时,为了在较低的误警率下提高关键词检出系统的检出率,提出了关键词恢复算法,通过对被拒绝的语音信号再次分析从而恢复关键词。实验表明:引入关键词恢复算法,在误警率为10%时,检出率从78.1%提高到85.3%;检出率的上限从85.0%提高到92.0%。采用关键词恢复的检出算法极大地改善了系统性能,不仅在较低误警率时检出率有很大的提高,而且关键词检出率的上限也得到很大的提高。
In order to improve the detection rate of the keyword detection system at a lower false alarm rate, a keyword recovery algorithm is proposed when the recognition performance is reduced due to replacement errors in the keyword detection system. The speech signal is analyzed again to recover the key words. Experiments show that when the false alarm rate is 10%, the detection rate is increased from 78.1% to 85.3% with the keyword recovery algorithm introduced. The upper limit of the detection rate is increased from 85.0% to 92.0%. The detection algorithm using keyword recovery greatly improves the system performance, not only greatly improves the detection rate at the lower false alarm rate, but also greatly increases the upper limit of the keyword detection rate.