logfbank functionstrange winstep size

Question

logfbank functionstrange winstep size

LeeYongHyeok opened this issue 6 years ago · comments

LeeYongHyeok commented 6 years ago

Hello, i will use this project, in multimodal speech recognition.

when i use this logfbank function in python3 env, i got 237 frames logfbank feature, using 1.19s wav file.

I check my option, winlen=0.02, winstep=0.01 but output frames are same.

I change the option to winlen=0.04, winstep=0.02, and i got 118 frames.

what's the matter?? i confuse ...

I installed by 'pip3 install python_speech_features' command, and my os is ubuntu.

thanks!

James Lyons · Answer 1 · Tue Jan 15 2019 17:00:30 GMT+0800 (China Standard Time)

Is it a stereo file? Try either combining the channels or sending in the channels separately

…

On Tue, 15 Jan 2019, 6:51 PM LeeYongHyeok ***@***.*** wrote: Hello, i will use this project, in multimodal speech recognition. when i use this logfbank function in python3 env, i got 237 frames logfbank feature, using 1.19s wav file. I check my option, winlen=0.02, winstep=0.01 but output frames are same. I change the option to winlen=0.04, winstep=0.02, and i got 118 frames. what's the matter?? i confuse ... I installed by 'pip3 install python_speech_features' command, and my os is ubuntu. thanks! — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#79>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABn1QceLPP1JzhXpJbSRoe6qC7m9I2uuks5vDZZ9gaJpZM4aAft8> .

LeeYongHyeok · Answer 2 · Tue Jan 15 2019 17:04:07 GMT+0800 (China Standard Time)

oh, i fix it thank!