近年来，随着神经网络在语音识别领域应用中的快速发展，深度学习被应用到声纹识别领域，取得了很好的效果。本文先是介绍了声纹识别的基本理论，说明了语音信号预处理和特征识别的一般方法，而后又介绍了一种基于LSTM神经网络的端对端声纹识别算法，从理论上说明了这种算法的优越性。通过这种算法构建的说话人声纹识别模型，大大节省了模型训练的时间，训练效果较好。 In recent years, with the rapid development of neural networks in the field of speech recognition applications, deep learning has been applied to the field of voiceprint recognition with good results. In this paper, we first introduce the basic theory of voiceprint recognition and illustrate the general methods of speech signal preprocessing and feature recognition, and then we introduce an end-to-end voiceprint recognition algorithm based on LSTM Neural Network to illustrate the theo-retical superiority of this algorithm. The speaker vocal pattern recognition model constructed by this algorithm greatly saves the time of model training and the training effect is better.