How about the perforemance on ASR task?

google-research / leaf-audio

LEAF is a learnable alternative to audio features such as mel-filterbanks, that can be initialized as an approximation of mel-filterbanks, and then be trained for the task at hand, while using a very small number of parameters.

Apache License 2.0

495 stars 52 forks source link

How about the perforemance on ASR task? #3

Open SoonSYJ opened 3 years ago

SoonSYJ commented 3 years ago

Hi, Thanks for your excellent work. Just asking about the performance on ASR task? Do you have any expriment result on LibreSpeech or any other corpurs?

SoonSYJ commented 3 years ago

Furthermore, how do you think about the latency introduced by the learnable front-end. As I tested, with batch_size 32, the time consumption is 0.02s(mel-filterbank) and 0.4s(LEAF) respectively.

peter-yh-wu commented 3 years ago

+1 awesome work, very interested in the ASR performance as well, thanks!