Hello Chia-Chun Fu, thank you for this article. It has been very helpful. I am currently facing a challenge with data preparation for a Tensor flow model ( vggish model with audioset data) in android. The challenge is mainly when it comes to calculating the MFCC in Android. In as much as I try to get similar MFCC values like those in tensorflow’s python implementation of MFCC ,all attempts haven't been successful. Can you advice on a way to overcome this hurdle?