摘要
针对传统的特征提取算法与分类识别模型对说话人年龄的识别精确度较低,且受外界噪声影响较大等问题,提出了一种结合改进型语谱图与自建DenseNets网络对性别与年龄识别的方法。首先制作语谱图数据集并改进语谱图的特征提取方式,然后搭建网络模型对语谱图进行分类识别,最后对算法模型的性能进行分析,并实现在线识别说话人的声纹特征。实验结果表明:该算法模型可以有效识别说话人的性别与年龄区间,具有一定的实用意义。
In view of the low accuracy of speaker age recognition by traditional feature extraction algorithms and classification recognition models,and the great influence of external noise,a method for gender and age recognition is proposed by combining the improved language spectrograph and self-built DenseNets network.Firstly,the spectrogram data set is made and the feature extraction method of spectrogram is improved.Then,a network model is built to classify and recognize the spectrogram.Finally,the performance of the algorithm model is analyzed and online recognition of the speaker’s voice print features is realized.The experimental results show that the proposed algorithm can effectively identify the speaker’s gender and age range,and has certain practical value.
作者
朱梦帆
汪志成
戴诗柏
ZHU Mengfan;WANG Zhicheng;DAI Shibai(School of Mechanical and Electronic Engineering,East China University of Technology,Nanchang 330013,China)
出处
《仪表技术》
2022年第1期66-70,73,共6页
Instrumentation Technology
关键词
语谱图
稠密卷积神经网络
性别识别
年龄识别
spectrogram
dense convolutional neural network
gender recognition
age recognition
