**** BEGIN LOGGING AT Sat Mar 17 03:00:00 2018 Mar 17 03:07:12 ds2: pocketsphinx for trigger word detection is not very accurate but I thought i could iron that out with some tuning Mar 17 03:07:24 another implementation for trigger words is through deep learning neural nets called snowboy, it is much better fared for our embedded application Mar 17 03:10:21 so for trigger word detection, I'll use snowboy. it has much better accuracy than pocketsphinx in this regards Mar 17 03:15:18 ds2: pocketsphinx is best suited for voice recognition because it is only platform which is dedicated to be used on embedded ARM processors and offers much more accuracy for the processing it uses Mar 17 03:17:19 it also has a huge language model, thus can be flexible/scalable in the variety of words detected Mar 17 03:20:18 other speech recognition models would't be able to work with full potential on the beagle Mar 17 03:24:31 I think the combination of Snowboy for trigger word detection and pocketsphinx for voice recognition would be the best suited for the offline smart speaker! Mar 17 03:37:03 snowboy is not open source Mar 17 03:37:19 apoorvtintin: what is your background? Mar 17 03:56:32 ds2: I am an undergraduate student with majors in electronics, I am interested/working in the field of signal and image processing Mar 17 03:56:47 and embedded systems Mar 17 03:57:52 ds2: i am sorry, i only have experience working pocketsphinx, if snowboy is not opensource it is out of question Mar 17 04:01:13 ds2: I have experience working with deep neural nets, if not snowboy, maybe i can create a neural system by mysel using gstreamer for fetching audio, I will make a detailed plan on this and consult you again Mar 17 04:06:24 apoorvtintin: I see. Wanted to put your comment in context :D Mar 17 04:06:52 snowboy is a bunch of closed source libraries plus there is no "free" any-speaker models Mar 17 04:08:12 we can address this part... what you need to do is run this by another mentor (all of us evaluate the proposals)... the proposals need to be convincing that you can handle the work in the GSoC time frame Mar 17 04:25:44 Any mentor online Mar 17 04:39:33 ds2: thanks, i'll do that :) Mar 17 11:17:58 hey, any mentor interested in "Remote Seismometer Node" project?? I very much want to work for BB for GSoC 18. But am having a hard time trying to discuss project ideas here. Mar 17 11:18:49 Seems like all the mentors are interested in kernel related ideas only Mar 17 11:24:01 Also I was able to get the thing described at https://www.sparkfun.com/tutorials/235 up and running with a Arduino Nano and adxl335 Mar 17 13:47:58 jkridner: I've shared the draft proposal. Request for feedback Mar 17 13:52:21 on elinux i hope. Mar 17 13:52:47 yes Mar 17 16:35:43 jkridner[m]: Have you sent out mentor invites from summerofcode.withgoogle.com yet? Mar 17 18:17:14 @jkrinder how are projects going to be selected ? on priority or will all listed ideas will be accepted ? Mar 17 18:22:15 Pac23: G allots slots. And thats the max number. Further, the proposal should look convincing to the mentors. Mar 17 18:22:30 G = Google Mar 18 01:13:30 Wormo: abhishek_[m]: ds2: jkridner[m]: zeekhuge: nerdboy: _av500_: First draft of my proposal: https://elinux.org/BeagleBoard/GSoC/BeagleBone-based-Serial-Terminal-Server . Please take a look and tell me if you have any feedback/suggestions **** ENDING LOGGING AT Sun Mar 18 03:00:03 2018