FCNN深度学习模型及其在动物语音识别中的应用

为解决使用语音信号准确识别动物以保护和研究野生动物的问题,提出一种全连接算法与稀疏连接算法相结合的全卷积神经网络(FCNN: Fully Convolutional Neural Network),用于语音的自动识别.利用全连接算法提取更多的组合特征,稀疏连接算法筛选重要特征可加快收敛速度.同时给出了具体的模型结构及算法流程,并进行了动物语音识别实验.实验结果表明,该全卷积神经网络深度学习算法是一种语音自动识别的有效方法,解决了蛙声识别问题,为动物语音识别提供参考.

FCNN Deep Learning Model and Its Application in Animal Speech Recognition

In order to solve the problem of using voice signals to accurately identify animals so as to protect and research wild animals. We propose a FCNN (Fully Convolutional Neural Network) combining a fully connected algorithm and a sparse connection algorithm for automatic speech recognition. The fully connected algorithm is used to extract more combined features, and the sparse connection algorithm to select important features to speed up the convergence. The specific model structure and algorithm flow are given, and speech recognition experiments are carried out. The experimental results show that the fully convolutional neural network deep learning algorithm is an effective method for automatic speech recognition. It can solve the problem of frog sound recognition and provide a reference for animal speech recognition.