Here I have a question about SILK:
The input speech will first pass the LTP filter, then LTP_res will pass the STP(LPC) filter to produce the residual LPC_res, then the LPC_res will be ready to noise shaping analysis.
This is my understanding of SILK, is it right or wrong?
If it is right, then why doesn't the LPC_res become the output of function:'void SKP_Silk_find_pred_coefs_FLP'