Voice deep neural network training method and device, storage medium and electronic device

• Main Authors: SHI HUIYU, OUYANG PENG
• Format: Patent
• Language: Chinese; English
• Published: 18.02.2022
• Subjects: ACOUSTICS; CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; MUSICAL INSTRUMENTS; PHYSICS; SPEECH ANALYSIS OR SYNTHESIS; SPEECH OR AUDIO CODING OR DECODING; SPEECH OR VOICE PROCESSING; SPEECH RECOGNITION

The invention discloses a voice deep neural network training method and device, a storage medium and an electronic device. The voice deep neural network training method comprises the steps of acquiring a mixed voice sample and a target sample voice, wherein the mixed voice sample comprises a target voice and a noise voice; inputting the mixed voice sample into a preset voice deep neural network model to obtain a predicted target voice, wherein the preset voice deep neural network model comprising an extractor, an enhancer and a reconstructor; and determining a preset voice deep neural network model as a target voice deep neural network model when a loss function determined by the target sample voice and the predicted target voice meets a preset condition, and based on the voice deep neural network comprising an extractor, an enhancer and a reconstructor trained in the scheme, and solving the problem that in the prior art, the target voice cannot be effectively separated from the mixed voice. 本发明公开了一种语音深度神经网络训