DSPRelated.com
Forums

Voice Compression for digital storage and playback

Started by Kunal June 18, 2004
I am making a Digital Dictation Machine and I need a solution to
digitize the voice, compress it and store it on a flash chip. It
should then also be able to decompress it and play it back through a
d/a. I need to control it using I2C.

Questions:
1) Which are the latest voice compression algorithms out there?. i
know of a-law u-law, but what about the ones used in cell phones and
other new ones.

2) I found "voice band audio processors" at ti, are these suited to my
application? Although they dont provide for external memory access,
could I use them for the compression atleast?

3)Another path would be to use a general purpose DSP chip and then
program it. This way i could get the peripheral support and still get
compression. Since this is a portable application, i plan to use the
5000 (low power)series. Could you help me narrow down this choice to a
specific chip or how should i estimate my processing needs.

4) any other useful information would be great:

Kunal
Kunal wrote:

> I am making a Digital Dictation Machine and I need a solution to > digitize the voice, compress it and store it on a flash chip. It > should then also be able to decompress it and play it back through a > d/a. I need to control it using I2C.
What Is I2C? I have much to learn.
> Questions: > 1) Which are the latest voice compression algorithms out there?. i > know of a-law u-law, but what about the ones used in cell phones and > other new ones.
You too have much to learn. A- and mu-law compress dynamic range, not storage requirements.
> 2) I found "voice band audio processors" at ti, are these suited to my > application? Although they dont provide for external memory access, > could I use them for the compression atleast?
That depends on what they process. What do their descriptions say?
> 3)Another path would be to use a general purpose DSP chip and then > program it. This way i could get the peripheral support and still get > compression. Since this is a portable application, i plan to use the > 5000 (low power)series. Could you help me narrow down this choice to a > specific chip or how should i estimate my processing needs.
If A programmable processor is "another path", what had you in mind for the first path? Will TI's routines be useful for that? What you want to do is part of what many telephone answering machines do. Perhaps you can find out how some of them do it. Jerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������
> > Questions: > > 1) Which are the latest voice compression algorithms out there?. i > > know of a-law u-law, but what about the ones used in cell phones and > > other new ones. > > You too have much to learn. A- and mu-law compress dynamic range, not > storage requirements. >
It is my understanding that a-law mu-law does reduce storage requirements by representing 16 bit samples in 8 bits. Not that it is very good compression :)
yo wrote:

>>>Questions: >>>1) Which are the latest voice compression algorithms out there?. i >>>know of a-law u-law, but what about the ones used in cell phones and >>>other new ones. >> >>You too have much to learn. A- and mu-law compress dynamic range, not >>storage requirements. >> > > It is my understanding that a-law mu-law does reduce storage requirements by > representing 16 bit samples in 8 bits. Not that it is very good compression > :)
I suppose that's a saving. Anyhow it stores 12 bits of dynamic range in 8 bits of storage space. Jerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������
"Jerry Avins" <jya@ieee.org> wrote in message
news:40d36d0c$0$3027$61fed72c@news.rcn.com...
> Kunal wrote: > > > I am making a Digital Dictation Machine and I need a solution to > > digitize the voice, compress it and store it on a flash chip. It > > should then also be able to decompress it and play it back through a > > d/a. I need to control it using I2C. > > What Is I2C? I have much to learn.
I2C = 'Inter-IC' - Philips's two wire bus for comms between ICs. Leon
The GSM vocoder gets voice down to 13 kbit/s (Search for 'Regular Pulse
Excitation - Long Term prediction'). ulaw/alaw are 64 kbit/s. Last I was
working with GSM, c.4 years ago, they (ITU-T) were talking about a 8 kbit/s
vocoder. I think 3G networks are using an adaptive multi-rate vocoder.
Have fun, Syms.


Leon Heller wrote:

> "Jerry Avins" <jya@ieee.org> wrote in message
...
>>What Is I2C? I have much to learn. > > > I2C = 'Inter-IC' - Philips's two wire bus for comms between ICs. > > Leon
Thanks. Jerry -- Engineering is the art of making what you want from things you can get. &#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;
A lot of the tiny MP3 players out there can also record/compress from
mic input. For a dictation machine you could code to as low as 16kb/s
with useful results. If you want to roll your own instead of hacking a
bought device, the MP3 player chips are easy to find and well
documented, and I believe Microchip and TI (at least) have apnotes
showing how to use their products in that application, and TI also
have MP3 (and other) libraries for their DSPs (ask them). There are
more efficient compression methods, but they generally take way more
code and horsepower for the compression process, and MP3 is way more
universal anyway.

Of course there are no/low-cost PC-based MP3 dictation machines out
there, but I assume you need to be independent of a PC or laptop?

On 18 Jun 2004 13:17:39 -0700, kunalshenoy@hotmail.com (Kunal) wrote:

>I am making a Digital Dictation Machine and I need a solution to >digitize the voice, compress it and store it on a flash chip. It >should then also be able to decompress it and play it back through a >d/a. I need to control it using I2C. > >Questions: >1) Which are the latest voice compression algorithms out there?. i >know of a-law u-law, but what about the ones used in cell phones and >other new ones. > >2) I found "voice band audio processors" at ti, are these suited to my >application? Although they dont provide for external memory access, >could I use them for the compression atleast? > >3)Another path would be to use a general purpose DSP chip and then >program it. This way i could get the peripheral support and still get >compression. Since this is a portable application, i plan to use the >5000 (low power)series. Could you help me narrow down this choice to a >specific chip or how should i estimate my processing needs. > >4) any other useful information would be great: > >Kunal
Tony (remove the "_" to reply by email)
"Jerry Avins" schrieb
> Leon Heller wrote:> > > > What Is I2C? I have much to learn. > > I2C = 'Inter-IC' - Philips's two wire bus for comms > > between ICs.
You might have an I2C bus in your PC, as this is IIRC used for temperature measurements (CPU, mainboard), perhaps as well for dynamic fan control. You might want to have a look at http://www.philipslogic.com/products/i2c/ Regards Martin
Symon wrote:
> The GSM vocoder gets voice down to 13 kbit/s (Search for 'Regular Pulse > Excitation - Long Term prediction'). ulaw/alaw are 64 kbit/s. Last I was > working with GSM, c.4 years ago, they (ITU-T) were talking about a 8 kbit/s > vocoder.
There's Enhanced Full Rate and Half Rate CODECs in GSM these days. Note that teh GSM codec does stuffliek comfort noise detection and generation which the OP may not desire (coder detects type of background noise, and decoder plays correct background noise when no speech bits are transmitted or received). Thomas