[Speex-dev] Echo cancellation - usage & porting to MIPS32 architecture

dimitar.belev at soneco.bg dimitar.belev at soneco.bg
Tue Apr 25 08:55:59 UTC 2017


Dear manager,

Our company is developing embedded product, which requires echo cancellation. We found the speex license is open source under GPL, so we can use the library source code.

I am struggling to port the echo cancellation part to MICROCHIP PIC32MZ family, which is based on M14K core from MIPS. The compiler is MICROCHIP XC32 C compiler.

Do you have kind of support for this project for porting to new architectures, and then add the already supported architecture to the main project.

I presume if you can offer support to port the library to MIPS32, then we can provide the project and you can include into the main branch.

Also if not possible, can we ask for some info during the porting, currently we have the following issues:

- There are many variables of “int” type. Is this type signed 32bit integer?
- Do you have some test data, how the echo cancellation algorithm should behave? It will be perfect if we can use it to debug the behavior of all algorithm blocks separately?
- In my project I use the onboard ADC to take samples from the MIC every 125uS (8KHz sampling rate) and PWM to output data to the speaker every 125uS. When I collect full frame from the Microphone, I pass trough the EC algorithm. Due to I need the speaker data of the same period, do I need to get earlier speaker data (for example 20-30 samples before the frame period in order the signal to be valid).

In general if you can help with info how to test and debug all blocks separately, we can port the entire EC algorithm and then I can send back the project, so you can include in the main branch, it will be tested and working.


Thank you in advance

Kind regards

Dimitar Belev
Soneco LTD
Mob: +359877398772
dimitar.belev at soneco.bg
www.soneco.bg


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.xiph.org/pipermail/speex-dev/attachments/20170425/2b559aa3/attachment.html>


More information about the Speex-dev mailing list