Deep learning acoustic models for human-device interaction in the edge

Go back

Finalized

Resources

The audio dataset will be provided.

Sorry! This challenge has ended.

Thematic

Research project aimed at students from any STEM-related field who wish to develop their TFG/TFM project on this challenge.

Introduction to the challenge

In human-device interaction, acoustic understanding is crucial for having fluent and natural communication.

In this context, edge AI helps enormously by reducing delays and carbon emissions. Additionally, it preserves human privacy as the information is processed directly in the device.

So, edge deep neural networks that allow understanding through the audio of what is happening in a concrete scenario are key elements in modern life.

Challenge posed

The research project objective is to generate a small-footprint neural network that uses audio input to help in human-device communication.

The project can be split into the following subtasks:

Exploration and pre-processing of an existing audio dataset
Use that data to train and evaluate a small-footprint neural network
Integrate the neural network in a constrained device