I used a data set with a sampling frequency of 16000hz, 1s for each voice, and two types of labels. I used mfcc for feature extraction and sent the trained model to the board.
In another project, I used another voice data set (sampling frequency is also 16000hz, two types, each voice 1s), and used the same model for training, but there was an error in the project:
ERROR: MFCC failed (-1004)
ERROR: Failed to run DSP process (-1004)
ERROR: Failed to run classifier (-5)
The error seems to occur when the device performs feature extraction(mfcc) on the voice. I think the parameters of the voice processing should be the same when the two data sets are used separately, but obviously this is not correct. After changing to the second data set, which parameters should be modified by the dsp module to solve this problem?
Thanks!
Hello @sherry,
Could you tell me which project are you using so I can have a deeper look please? I see different projects under your account.
I think the error code MFCC failed (-1004) is an “Out of dynamic memory error”.
Which board are you using?
Regards,
Louis
Hi @sherry -1004 is actually EIDSP_MATRIX_SIZE_MISMATCH - so one of the buffers passed into the function is the wrong size. If you could also share how you’re calling the model that would help (in addition to the project ID).