Auris helps the deaf and hard-of-hearing remain independent. Using an SVM classifier, Auris notifies users about peculiar sound events. Implemented as both a desktop and mobile application, our solution is scalable to low-powered, stand-alone devices.

The classifier was trained using 9,000+ audio clips from's public dataset, annotated into 41 distinct classes. This was sufficient for detecting a variety of important sounds, including knocking and alarms. Google Cloud Compute Engine was used to accelerate the training process. The classifier utilized 13 distinct sets of audio features extracted using pyAudioAnalysis.

Built With

Share this project: