When using pesq_batch the user might be using batched audio waveforms that originally had different length and were padded to match their length before creating the batch. The padding however changes the results.
This PR solves this by adding an extra optional argument lengths to pesq_batch. lengths should contain the original lengths of the signals before padding and batching.
When using
pesq_batch
the user might be using batched audio waveforms that originally had different length and were padded to match their length before creating the batch. The padding however changes the results.This PR solves this by adding an extra optional argument
lengths
topesq_batch
.lengths
should contain the original lengths of the signals before padding and batching.