Problem with extracting MFCCs into json file

[sloww]

Hi everyone,

I am working on developing a Vocal Detection System. I am extracting MFCCs from the audio (wav file) to prepare input dataset and labels for training a Vocal Detection System with CNN. However, after I formed a vocal time stamp list (including the starting time and ending time of the lyrics in the song) and try passing the (starting time and ending time) into the librosa.feature.mfcc to extract the MFCCs within a specific time interval, the program failed after running. I have tried to add a print function to track the extracted MFCCs which has a good result.

Any idea what would be the reason?

[Brian McFee]

Looks to me like you're trying to compute features on an empty buffer.

[sloww]

Thank You for your help with my problem! But I don't understand what is the empty buffer. Is it used to save the features? So what I am trying to do is to save the MFCCs and their labels (0 for non-vocal) and (1 for vocal) into a dictionary. I have added my code for doing the Feature Extraction Process and saving them into a dictionary below.