Re: Speech compression in dictation machines | Speech Coding

Technical discussions related to Speech Coding (all itu and other vocoders, ACELP, CELP, AMR, etc)

Dear All, I am looking for information on the compression done in portable dictation machines, such as the Olympus DS-3000. The files have an extension (DSS-Digital Speech Standard, file format from IVA-International Voice Association : Grundig, Philips and Olympus). From commercial documentation around, I could gather that the input is 11025 Hz, 16 bits, mono (176.4 kbits/sec). The compression is supposed to be optimized for further recognition. In SP mode the bit rate is about 14-15 kbps (compression factor of ~12). In LP mode the bit rate is about 6-7 kbps (compression factor 25). The dictation machine comes with a DSS player, which can play and convert (decompress) files to .wav. The decompressed files have reasonable quality, considering the bit rate. The files have to be converted to .wav before recognition by dictation software (e.g. IBM ViaVoice or L&H VoiceXpress). Any information or pointer to information is welcomed, particularly: - I would like to verify the sampling rate and number of bits of the input signal (before compression), type of compression algorithm used, and the exact compression factor. - I also would like to know about the DSS file format. - There exists an explanation on how to decompress .dss files into e.g. linear PCM (I mean the algorithm for decompression)? - I have found software that converts .dss to .wav. There exist any that converts .wav into .dss? Thanks and regards, Sara

(You need to be a member of speechcoding -- send a blank email to speechcoding-subscribe@yahoogroups.com )

Assuming you've had no luck with the obvious approaches (web searches for "dss codec", looking for linux drivers etc), I think your first line of attack should be to look at the machine and its manuals and packaging for patent numbers. If you can't find them there, do a patent search. Look up the patents (www.uspto.gov) and read thoroughly. Its amazing how much information you can sometimes glean that way. That should give you a good idea of the basic algorithm, or at least narrow the field. There are a few it could be. If its ADPCM, you've also got to guess the adaptation algorithm and quantisation tables - not at all easy. Good luck!

(You need to be a member of speechcoding -- send a blank email to speechcoding-subscribe@yahoogroups.com )

Sign in

Search speechcoding

Subscribe to speechcoding

speechcoding by Keywords

Discussion Groups

Discussion Groups | Speech Coding | Re: Speech compression in dictation machines

Speech compression in dictation machines - Author Unknown - Nov 20 10:28:00 2001

Re: Speech compression in dictation machines - Author Unknown - Nov 21 10:45:00 2001