摘要
针对佤语语谱图的识别无需考虑清、浊音的影响这一特征.利用傅里叶变换将佤语转换为对应的语谱图信息,将深度卷积神经网络的AlexNet模型用于佤语语谱图识别.实验表明,语谱图识别可以有效解决语音识别过程中清、浊音对实验识别结果的干扰,实验准确率达到96%.
Since it is unnecessary to consider the influence of clear and voiced sounds in the recognition of the Wa longuage spectrogram,this research uses the Fourier transform to transform the Wa language into the corresponding spectrum-related information.In addition,the AlexNet model of the deep convolutional neural network is introduced into the spectrum-based recognition of the Wa language.The result shows that the spectrum-based recognition of the Wa language can effectively solve the interference of clear and voiced speech in the process of speech recognition,and the experimental accuracy rate reaches 96%.
作者
王翠
王璐
解雪琴
和丽华
潘文林
WANG Cui;WANG Lu;XIE Xue-qin;HE Li-hua;PAN Wen-lin(School of Mathematics and Computer Science, Yunnan Minzu University, Kunming 650500, China)
出处
《云南民族大学学报(自然科学版)》
CAS
2019年第4期377-381,共5页
Journal of Yunnan Minzu University:Natural Sciences Edition
基金
国家自然科学基金(61363022)
云南民族大学研究生创新基金(2018YJCXS227)