半监督学习对十个口述数字的识别

Syllable Recognition Based on Semi-supervised Learning

提出了一种基于半监督学习的音节识别方法,该方法可在训练样本很少的情况下得到较好的识别效果。实验结果表明,在小样本情况下,半监督学习方法比传统的HMM方法具有更高的识别率。

A syllable recognition approach based on semi-supervised learning is proposed. This approach has an advantege of obtaining effective recognition when training data set are incomplete Experiment results prove that semisupervised learning is more efficient than traditional HMM in syllable recognition.