A pre-trained sound event detection neural network

This is a network trained earlier on clean data. It does not give very good results but is enough to show the system working. It can achieve above 90% on clean sounds and probably about 80% accuracy in 0dB SNR.<br>The network was saved manually (using the MATLAB 'save' command) after running the training code, and before running the testing code.本神经网络为此前基于纯净数据训练得到的模型。其性能未达最优，但足以验证系统可正常运行。针对纯净音频，该模型的识别准确率可达90%以上；在0dB信噪比场景下，准确率约为80%。该网络在训练代码执行完毕后、测试代码运行前，通过MATLAB的'save'命令完成手动保存。