本文选择上下文相关的三音子 (TriPhone)作为基本识别单元,实际设计中TriPhone的HMM模型采用无跨越式从左到右的结构,状态转移只能从左到右,包含3个发射状态和一个退出...
基于28个网页-相关网页
A crucial issue in triphone-based continuous speech recognition is the large number of parameters to be estimated against the limited availability of training data.
基于三音子连续语音识别的一个关键问题是在有限训练数据的条件下对大量声学模型参数的鲁棒性估计。
参考来源 - 连续语音识别中声学建模的组合聚类算法研究 in C·2,447,543篇论文数据,部分数据来源于NoteExpress
在保证稀有三音子全部被选择出的前提条件下,使三音子尽量平衡。
In the precondition of all rare triphones are selected out of raw materials, we tried to balance the triphone.
实验表明:基于语音学分类的三音子单元对识别性能有明显的改善,系统的首选误识率相对基线系统降低了28%。
Experimental results show that phonetic classification based on the triphone can greatly improve system performance. The proposed method reduces the error rate by 28% compared with a baseline system.
当然,它是小调的,人声将会进入,但低音是这样,按基本的二拍子缓慢地行进,但在那下面还用了一个三连音的节奏细分。
Of course, it's in minor and the voices will come in, but the bass is going, sort of plodding along in a basic duple but with a triple subdivision underneath of that.
应用推荐