Hello, is there any particular reason why preemphasis is set explicitly to zero in the mfcc calculation of the processAudioWindow c++ function instead of 0.97 which is what is used in the preprocessing and DS_AudioToInputVector?
Because we’re doing streaming, we need to do preemphasis ourselves (see StreamingState::feedAudioContent) to avoid a discontinuation between the last sample of the previous window and the first sample of the current window.