Speech Quality Feature Enhancement Analysis Based on Joint Loss Function
In order to ensure that the cost function can carry out the analysis process according to the character-istics of human ear perception,a speech enhancement deep learning algorithm combined with loss function is de-signed on the basis of the above.The auditory data of human ear are added to the calculation process of loss func-tion.The results show that the hybrid loss function designed in this paper can achieve obvious optimization of en-hanced speech quality.When the conventional loss function MES is used,the enhanced speech is obviously distort-ed.When processing with joint loss function,the degree of enhanced speech distortion can be significantly reduced and better speech quality can be obtained.
speech enhancementjoint loss functionhearingdistortion of speech
NETL
NSTL
万方数据
维普
