Need help in DSP

[Al]

I have a project in mind which would require that I identify an object
by the sound it emits. It will be the same object all the time and it
has a distinctive audio characteristic. I am fairly certain that the
audio signature is distinctive.

I would like to measure the amplitude of at least three frequencies,
more is better. If there is a match within a specified window, I will
have identified the object.

I know I could do this by making an analog notch filter for each
frequnecy, doing an A/D conversion, and comparing against a template.
But I don't have much room and the power requirement is strict, i.e.,
field operation for weeks at a time.

Has anyone had experience with DSP chips? I would appreciate a
recommendation for one that could be programmed in C. I don't have any
DSP experience, so something that would not be too difficult to learn
would be nice.

Thanks in advance.

Al

--
There's never enough time to do it right the first time.......

 

How about some more details:
-> frequency range of interest
-> required frequency and ADC resolution,
leading to MIPS requirement for FFT
-> how often does the device have to be awake?
-> size/price requirement
-> quantities to be made, just one or 1million?

On Sun, 04 Apr 2004 15:54:11 GMT, Al <no.spam@here.com> wrote:

I have a project in mind which would require that I identify an object
by the sound it emits. It will be the same object all the time and it
has a distinctive audio characteristic. I am fairly certain that the
audio signature is distinctive.

I would like to measure the amplitude of at least three frequencies,
more is better. If there is a match within a specified window, I will
have identified the object.

I know I could do this by making an analog notch filter for each
frequnecy, doing an A/D conversion, and comparing against a template.
But I don't have much room and the power requirement is strict, i.e.,
field operation for weeks at a time.

Has anyone had experience with DSP chips? I would appreciate a
recommendation for one that could be programmed in C. I don't have any
DSP experience, so something that would not be too difficult to learn
would be nice.
Al

 

[James Meyer]

On Sun, 04 Apr 2004 15:54:11 GMT, Al <no.spam@here.com> posted this:

I have a project in mind which would require that I identify an object
by the sound it emits. It will be the same object all the time and it
has a distinctive audio characteristic. I am fairly certain that the
audio signature is distinctive.

You don't need a DSP for that job. A micro-power PIC chip should work
just fine. Use the on-chip ADC capability and do a series of FFT conversions to
get the information you need. A DSP would only be necessary if you needed to do
a "real time" continuous detection of something with a much higher data rate.

Jim

 

[Al]

In article <40704597.1742315@news2.news.adelphia.net>,
g9u5dd43_nospam@yahoo.com wrote:

How about some more details:
-> frequency range of interest

100 to 3000 Hz

-> required frequency and ADC resolution,

8 bits should be enough

leading to MIPS requirement for FFT

-> how often does the device have to be awake?

the detector would be awake all the time, but the rest of the circuitry
would have to come awake when the object was detected. Maybe a two
chips, one that is on all the time would have low power requirements and
the other could be higher. After detection, full power for 20 seconds or
so would be enough time.

-> size/price requirement

< $50, eight cubic inches less power source

-> quantities to be made, just one or 1million?

proof of concept - maybe a dozen, later? who knows?

On Sun, 04 Apr 2004 15:54:11 GMT, Al <no.spam@here.com> wrote:
I have a project in mind which would require that I identify an object
by the sound it emits. It will be the same object all the time and it
has a distinctive audio characteristic. I am fairly certain that the
audio signature is distinctive.

I would like to measure the amplitude of at least three frequencies,
more is better. If there is a match within a specified window, I will
have identified the object.

I know I could do this by making an analog notch filter for each
frequnecy, doing an A/D conversion, and comparing against a template.
But I don't have much room and the power requirement is strict, i.e.,
field operation for weeks at a time.

Has anyone had experience with DSP chips? I would appreciate a
recommendation for one that could be programmed in C. I don't have any
DSP experience, so something that would not be too difficult to learn
would be nice.
Al

--
There's never enough time to do it right the first time.......