Timbre Dataset Generation

The proposed model uses the timbral properties of voice, that is hardly used in any other research endeavors. The model is tested against a real-world continuous stream of audio, without any modification.

所提出的模型利用了语音的音色特征——这一特征在其他研究工作中极少被采用。该模型针对真实世界的连续音频流进行了测试，且未做任何修改。