would anyone in the DSPRelated community be interested and inspired to give a talk, possibly titled 'Essential DSP Knowledge for Embedded Systems Engineers' at the upcoming Embedded Online Conference? Possible topics that could be covered in the talk (just brainstorming here):
- basic filter design
- algorithm optimization
- leveraging core libraries
- hardware optimizations
- selecting DSP hardware
If you are interested in giving such a talk, please reach out directly to me through my email address (stephane) @ the domain name of this website.
If you are not interested but have suggestions as to what a DSP talk targeted at Embedded Systems engineers should cover, please reply in this thread. Thank you!
fixed point or float?
I've been working on real-time ASR (automatic speech recognition), which although codebases such as Kaldi offer a RTF > 1, still pose challenges for real-time applications incorporating signal processing, as the ASR RTF numbers typically apply over 1 sec or longer, and not at small intervals (such as 10-30 msec often used in voice/telecom systems). These engineering and software architecture obstacles apply to anything from Raspberry PI up to x86 rack servers in data centers.
Would a talk on this fit ? Maybe under "core libraries" if we can consider Kaldi ASR as a set of core libs, having been around since 2009.
Thanks for proposing to present about real-time ASR. I think this could be a very good topic for the upcoming 2021 DSP Online Conference in September and I cross my fingers that you'll still be interested in giving the talk by then.
For the Embedded Online Conference, what we are really looking for is an overview talk that would touch on many basic DSP concepts with a focus on an Embedded Systems Engineers audience.
Yea but ASR is not DSP. It's deep learning, a subset of machine learning.