Anyone know of a low cost speech recognition solution, it needs to be able to recognise commands preferably without voice training of around 30 words.
Phil
Speech Recognition
Feb 15, 2004
9 Replies
I suspect that "low cost" and "without training" may be mutually exclusive.
You might want to check over on the comp.dsp group.
Rich Webb Norfolk, VA
"Rich Webb" wrote
I suspect you are right :(
checked it and left, thanks anyway. The best I can find is Sensory Inc RSC 4128 was hoping someone here might know an alternative.
Phil
30 words *with* training is quite easy with a low-cost system, using a bank of band-pass filters. We were doing it on early machines like the TRS-80 25 years ago with add-on hardware. It could be done on a single-chip MCU now, using DSP.
Leon
Leon Heller, G1HSM
Email: aqzf13@dsl.pipex.com
My low-cost Philips LPC210x ARM development system:
http://www.geocities.com/leon_heller/lpc2104.html
"low cost" is relative - you'll have to define what that means ( or buy a macintosh :-) ). - RM
And before that, there was the "Housemaster" (IIRC) add-in board for Digital Group Z80 systems. Had voice control of several relays but it did require training.
One nifty aspect was that you could see the received pattern as it ran as well as after it was stored. It only sampled in two bands but it actually worked pretty well.
Rich Webb Norfolk, VA
Freeware, of course with source, and it has to run on a PIC or AVR, right ?
Rene
Ing.Buero R.Tschaggelar - http://www.ibrtses.com
& commercial newsgroups - http://www.talkto.net
TinyVoice, Circuit Cellar, Feb 1998. A motorola microcontroller system, with full assembler source code.
formatting link
I've played with the Sensory Voice Extreme kit a bit, and it works not too bad. You really should consider using training though, or else your accuracy will not be very high.
Sensory requires that you have them build the recognition files for speaker independant recognition, which is a bit of a pain. But, it is easy to use in training mode.
Mike Anton
Hello Phil, i hope some interest for you is device SpeechBox, it's mine release of Sakrament.SpeechRecognition Engine for TMS...5402,
formatting link
Program is ANSY C, it's possible to deliver as file '.obj' for you processor. 2 models are released: HMM with teaching on PC & 'dinamical programming' with teaching one time pass for each word, the last has more high probability of error. It's without teaching (for microprocessor's system) HMM for phonems recognition, speed & safety are low.
Join the Discussion
Have something to add? Share your thoughts — no account required.
Didn't find your answer?
Ask the community — no account required