介绍了利用三音子模型和基频信息提高汉语连续数字串的识别率。
Research work on the influence of training method in mandarin connected digit speech recognition;
各个音节的基频是随时间而变化的，其不同轨迹构成了声调，利用基频信息进一步提高了对汉语连续数字串的识别率。
In Chinese the pitch of syllable varies with time, whose track forms different tones. By using the pitch information the error rate of Mandarin connected digits is further decreased.
应用推荐