tejas-shastha / audio_emergency_nonros

Recognise user emergency signal from audio input

Audio emregency

User hums to create an audio emergency signal

Procedure

Run python emerg_record.py <name> to record a 3s clip of the user humming. This generates .wav in the output directory and subsequently calls the evaluation script on this wav file
Optionally run python emerg_evaluate.py <name> to process .wav file in output directory and recognize the dominant frequencies.
Either process generates _freqs.csv in output directory containing dominamt frequencies.
Run python emerg_test.py <name> to start up the live testing process. This requires a _freqs.csv in output directory with correct dominant frequencies. Output is cluttered with ALSA trash messages, ignore it. In another terminal, run tail -f result.csv to get a cleaner output.
If hum is recognised correctly, you see a string "hum" at the end of the output line. If talking is recognised, you see a string "talk".

TODO

1. Include everything into ROS
2. Have a topic publish info and read that instead of tail -f result.csv

Investigate

1. Why batch processing gives different results between realtime and wave file execution