Hello. Earlier I asked if it would be possible to use one common large codebook for all audiofiles. Can you estimate how the following system would work? 1. The audio would be divided to 16 samples blocks. 2. For each block, the first sample would be stored without any encoding. 3. Linear interpolations between the stored samples are computed. 4. The residues (length of 16 samples) between the original audio and the linear interpolation are computed. Residues would go from 0.0 to 0.0. 5. The residues are vector quantized. For each 16 samples, one sample and one codebook reference would be stored. With 16-bit audio it would mean 1:4 compression ratio. Comments? Juhana --- >8 ---- List archives: http://www.xiph.org/archives/ Ogg project homepage: http://www.xiph.org/ogg/ To unsubscribe from this list, send a message to 'vorbis-dev-request@xiph.org' containing only the word 'unsubscribe' in the body. No subject is needed. Unsubscribe messages sent to the list will be ignored/filtered.
At 6:51 pm +0300 17/6/01, Juhana Sadeharju wrote:>Hello. > >Earlier I asked if it would be possible to use one common large >codebook for all audiofiles. Can you estimate how the following >system would work? > >1. The audio would be divided to 16 samples blocks. >2. For each block, the first sample would be stored without any encoding. >3. Linear interpolations between the stored samples are computed. >4. The residues (length of 16 samples) between the original audio and >the linear interpolation are computed. Residues would go from 0.0 to 0.0. >5. The residues are vector quantized. > >For each 16 samples, one sample and one codebook reference would be stored. >With 16-bit audio it would mean 1:4 compression ratio. > >Comments?Is this going to give results significantly better than IMA-ADPCM, which gives a similar compression ratio for a fairly low decode overhead and simpler decode loop? --- >8 ---- List archives: http://www.xiph.org/archives/ Ogg project homepage: http://www.xiph.org/ogg/ To unsubscribe from this list, send a message to 'vorbis-dev-request@xiph.org' containing only the word 'unsubscribe' in the body. No subject is needed. Unsubscribe messages sent to the list will be ignored/filtered.
>Is this going to give results significantly better than IMA-ADPCM, >which gives a similar compression ratio for a fairly low decode >overhead and simpler decode loop?Also the quality matters, but I have no idea. I suppose a VQ method outperforms the adpcm method because VQ keeps an external codebook but adpcm just reduced the total number of bytes. How IMA-ADPCM works? Documentation? What would be the error signal? Are there any source code? Sox could not decode IMA-ADPCM files I have found from one game. Could someone of you tell me how to use the VQ code of Vorbis to make the codebook for the simple system I proposed? Juhana --- >8 ---- List archives: http://www.xiph.org/archives/ Ogg project homepage: http://www.xiph.org/ogg/ To unsubscribe from this list, send a message to 'vorbis-dev-request@xiph.org' containing only the word 'unsubscribe' in the body. No subject is needed. Unsubscribe messages sent to the list will be ignored/filtered.