How to get MFCC feature when I do not want to use tensorflow?

[iamweiweishi]

Thank you for your geat work. I trained a TCResNet8 model and then converted to tflite-model successfully. But I do not know how to use the tflite model when I do not use tensorflow to extract the MFCC feature of wavs. I tried to use 'python_speech_features' to extract MFCC, but it seems like the extracted feature is totally different with the same input wav. Could you please give me some tips on this? Thanks a lot.