Hearham uploader software
2GB free disk space Any recent Ubuntu/debian should work. SDR device for input is recommended.
Run "./setup.sh" once to download and compile all necessary files. May take 10min or so.
Run ./HearHamUploader.py to start the recognizer.
hear.config has some options you may want to try, there are many options in the referenced .jconf files too.
With a radio next to the mic (or better yet, SDR software and the audio loopback described here: https://www.youtube.com/watch?v=RSeINGM68A8) you should see text roughly equivalent to what is spoken. When it's set up, enter your secret upload key that you got when setting up a channel on hearham.com. It should be noted that the server side software does take into account almost matching and phonetics - Charlie Oscar Oscar Lima!
The system in its current form tells what words you are saying but has a hard time distinguishing letters especially, and anything with static, as is common on the radio. There are various other systems that would be worth a look for anyone researching speech-to-text:
- Mozilla DeepSpeech
- Kaldi
- GoVivaci
- Google Cloud
- Houndify
- IBM Watson
- Microsoft Azure
and others?? That's just a list based on Mycroft's, some paid and some free.