Inspiration
America is built on immigrants. Individuals who span from all around the world, bringing diverse cultures, experiences and language barriers. Despite the many things that can divide us, we are grounded by countless similar events such as natural disasters. First Response is a system that interprets the language spoken by the person facing impeding danger whether that be: flooding, hurricane, fire, or any sort of disaster. Analyses the language in a speech to text API, then using ML responds to the individual in their native language with key safety precautions that can be taken to evade the impeding danger. The system would inform the person of the predicted weather patterns based on real time news-weather feed. We hope to expand on this idea by including: optimum evacuation routes and nearby hosts who are willing to host evacuated individuals.
We decided to implement a hardware accelerated circuits to attempt to mitigate and shorten the training time as well as processing time for machine learning and did so by doing thorough data cleaning of our inputs such as a band pass filter made of active components to limit low and high frequency noise, a fourier transform to split the sound waves into its corresponding sine waves and then parallelizing the layers of our network to allow simultaneous operation. Our ML attempts to recognize the language being spoken using a small amount of training data with a convolutional network which is trained on spectro-graphs generated from our audio input.
Log in or sign up for Devpost to join the conversation.