looking for audio (compression) project

Are you interested in a purely software project or more of a hardware/ software oriented project?

More on the hardware side you might like to say implement a massively parallel MP3 audio processing system on an FPGA. E.g. have a separate processor soft core for each MP3 channel, and provide a means to have some sort of user defined processing done on each MP3 encoded stream. i.e. MP3 encoding + processing or some sort (filters, effects, whatever). Perhaps make use of C-to-hardware compilers to convert some of the C code to hardware to improve performance.

Don't just think big though, small can be good too. How about an MP3 encoder/decoder using a tiny 8 pin PIC or something. Hand optimised assembler, single chip solution = neat.

Dave.

Do they make 100MHz PIC?

They have already suggested 1001 algorithm for the quality audio compression. All of those algorithms offer somewhat 30% better compression for the same quality compared to MP3, for the cost of the several times more of computing burden. All of those algorithms are based on either ATC or forward/backward LPC with vector coding. This area is also one of the most heavily patented, so you can't write a line of code without infringing somebody's intellectual rights.

What for? The main advantage of MP3 is that it is widely accepted. The

30% gain is not worth incompatibility and increased computing needs.

Find the different area for your enthusiasm.

Vladimir Vassilevsky DSP and Mixed Signal Design Consultant

The latest AAC implementation already offers a 75% quality improvement over MP3 at the same bitrate and is probably the final audio format we'll ever see. On the other hand, people thought the same about MP3

10 years ago, so maybe I'm just a narrow-minded prick. But it's finalized in terms of the concept it uses to compress audio, which is just an extended version of MP3, and it won't improve beyond that.

You need to focus on a concept that intelligently recognizes and discerns different elements in the audio and encodes them efficiently (hint: MIDI) For singing, the application analyzes the primary characteristics of the person's voice: the resonance, pitch etc. stores that in the header followed by instructions on the words he speaks, the volume, the length, whatever. It should be prevalent in about 20 years.

y e

lan

nce

!

I'd concentrate on some other end of the market, like low bit rate speech encode/decode. I think the present day solutions are piss poor at best. The phone and digital radio companies have spent fortunes on such technology, but with low latency in mind since they have to work real time. A voice encode/decode scheme designed for archive (latency be damned!) would be useful. There are all sorts of voice applications that are stored for eons to cover one's butt, such as public service telephony, stock trades, digital black boxes in aircraft, and probably NSA taps.

Senix do (or did) with the SX processor.

PIC was just an example, the idea being to get a MP3 encoder or decoder in the minimum hardware possible. Or how about an MP3 decoder using 7400 series logic? :->

Dave.