Inspiration
Fun
What it does
Uses machine learning algorithm to recognize sounds from different objects in real time.
How I built it
We modified the pyAudioAnalysis library to fit our needs and used its SVM machine learning algorithm to classify sounds from different objects. We had to use multi-threading in order to get real-time audio data and have it analyzed at the same time to reduce latency between the signal and dentition of the signal.
Challenges I ran into
The first challenge we ran into was accounting for the amount of ambient noise we had in our room. We did this classifying noise into its own category.
The largest challenge was to modify the pyAudioAnalysis library to take real-time data instead of saved ".wav" files.
Finally, the last challenge was to get different processes running on different threads to reduce latency.
Accomplishments that I'm proud of
Achieved 90% accuracy in recognizing objects with sound alone.
What I learned
multi-threading machine learning signal processing python
What's next for Object Recognition with Sound Waves and AI
In the future we will attempt to decrease the latency and use it to different things such as play real-time music, and play games.
Built With
- machine-learning
- pyaudioanalysis
- python
Log in or sign up for Devpost to join the conversation.