How to generate speech signal from spectral envelope, aperiodicity, fundamental frequency, V/UV signal

I have implemented the network as shown in fig which takes 2 inputs namely, video input and mfcc(audio) input. Video input consists of lip images and audio input is mfcc of corresponding video frame. The video and mfcc frames are passed through several layers and then added to generate speeech parameters. I have found fundamental frequency, spectral envelope, V/UV speech, fundamental frequency. I have taken ifft of spectral envelope to generate sound but it generates random signal.

please guide how to generate speech signal from speech parameters.

[Arun]

Could you share these parameters data, so that I can try to generate results at my end.

[Shilpa Sonawane]

ok sir