论文部分内容阅读
语音作为传递信息的一种常用手段,在人们的日常生活中有着非常重要的地位。随着科学的发展,语音识别愈来愈受到人们的重视。本文提出一种基于流形学习的特征提取方法——邻域保持嵌入(NPE)算法用于语音识别领域。流形学习是近几十年发展起来的降维方法,在图像识别领域已有应用,但在语音识别领域的应用非常之少。实验结果表明该算法可取得较好的识别率,同时所提取的特征稳定,计算速度快。
As a common means of transmitting information, voice plays a very important role in people’s daily life. With the development of science, speech recognition has drawn more and more attention. In this paper, a new feature extraction method based on manifold learning-neighborhood preserving embedding (NPE) algorithm is proposed in the field of speech recognition. Manifold learning is a dimension reduction method developed in recent decades. It has been applied in the field of image recognition, but its application in speech recognition is very rare. Experimental results show that the algorithm can achieve better recognition rate, while the extracted features are stable and computational speed is fast.