Sound Recognition

Inspiration

I want a good offline recognition tool. This is part of Human–computer interaction (HCI).

What it does

Work In Progress. Recognises limited number of sounds. Currently training on only 8 short worlds. Curretly the output is just an "debug" text file.

How I built it

It is not made publicly. I run directly from the source code. Currently just feeding WAVE files. pyaudio for audio processing.

Challenges I ran into

Speed, Capacity. Speed on running, speed of training. To feed one row of record it takes longer then one step. One of the problems I'm using an Jason Carver implementation of HTM which is not using numpy also I had modified some parts of code. Pickle is a problem it takes a lot of memory save/load, recursive and mutable were crashing python.

Accomplishments that I'm proud of

More bugs that I had introduced.

What I learned

Other

No ETA on this project, I'm very slow.

What's next for Sound Recognition

More testing. After succeeding in this I will go to Vision.

Built With

• python