xiph
Processing of 8KHz / 16KHz Sample rate audio signals
though this question was posted earlier in the discussion threads, I could not see a definitive answer on the same. I would like to understand if the RNNoise Demo can be used for 8KHz / 16KHz PCM samples.
is the training depends on the sample rates?
can anyone suggest what changes I should perform in order to adapt the decoder to process 8KHz/16Khz sample rate speech
@amadupu I think the RNN must do train again, assumed the sample rate was different, or...we have to change the some hypeparameter otherwise?
yeah..looks like retraining is required...but i feel there is more to it than retrainng...esp with signal processing that needs to be done for lower sampling rates..
@venkat-kittu try considering upsampling if that suits to your scope of work
regards arun
On Wed, Aug 1, 2018 at 6:13 PM, Loreto Parisi [email protected] wrote:
@amadupu https://github.com/amadupu I think the RNN must do train again, assumed the sample rate was different, or...we have to change the some hypeparameter otherwise?
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/xiph/rnnoise/issues/31#issuecomment-409561350, or mute the thread https://github.com/notifications/unsubscribe-auth/ANl1xP-G4gGzjrDtKNkvGqLt4TxEdY6cks5uMaJfgaJpZM4VDm6w .
If running at 8kHz can save some computing power, it would be significant when run in parallel.
yeah..looks like retraining is required...but i feel there is more to it than retrainng...esp with signal processing that needs to be done for lower sampling rates.. @venkat-kittu try considering upsampling if that suits to your scope of work regards arun … On Wed, Aug 1, 2018 at 6:13 PM, Loreto Parisi @.***> wrote: @amadupu https://github.com/amadupu I think the RNN must do train again, assumed the sample rate was different, or...we have to change the some hypeparameter otherwise? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#31 (comment)>, or mute the thread https://github.com/notifications/unsubscribe-auth/ANl1xP-G4gGzjrDtKNkvGqLt4TxEdY6cks5uMaJfgaJpZM4VDm6w .
Many pipelines that would require upsampling and then downsampling again because the rest of the pipeline is already running at 8000 and would be really inefficient.
