论文部分内容阅读
在汉语数码语音识别（ Ｍ Ｄ Ｓ Ｒ）中，“２”和“８”是最易混淆的一对语音。文章分析了“２”和“８”混淆的原因，发现可用于分辨“２”和“８”的区别特征在于其共振峰轨迹的差异。因此文章提出了基于共振峰轨迹的判决算法（ Ｆ Ｔ Ｂ Ｄ）来分辨“２”和“８”。实验表明，使用 Ｆ Ｔ Ｂ Ｄ 算法，使 Ｍ Ｄ Ｓ Ｒ识别率从９６０％ 提高到 ９７７％ ，“２”和“８”的识别率从 ９１％ 提高到９９％ ，消除了这对语音的混淆，提高了 Ｍ Ｄ Ｓ Ｒ 的整体性能
In Chinese Digital Speech Recognition (MDSR), “2” and “8” are the most confusing pairs of voices. The article analyzes the reasons for the confusion between “2” and “8” and found that the distinctions that can be used to distinguish “2” and “8” are characterized by differences in their formant trajectories. Therefore, the paper proposes a formant algorithm based on formant loci (F T B D) to distinguish “2” and “8”. Experiments show that using F T B D algorithm, the recognition rate of M D S R is improved from 96.0% to 97.7%, and the recognition rate of “2” and “8” is increased from 91% to 99% Obfuscation of speech improves the overall performance of MDSR