Evaluation Method of Timbre Thickness for Speech Based on CNN
Zhongwen Tu, Baoyu Cao, Lihao Wang, Kemin Wang, Wen Liu · 2019
In order to evaluate the thickness of speech timbre, this paper uses the spectrogram as the data input, and uses caffe model to study the thickness evaluation of speech timbre with AlexNet and GoogleNet as the neural network architecture respectively. The experimental results show that CNN has high accuracy in the field of speech timbre evaluation, and can effectively identify the thickness of speech timbre.