From its presentation in September 2020, the CIDAI initiative has been working to promote applied AI projects and tech transfer in Catalonia. As one of its founding members, the i2CAT Foundation is deeply committed to accomplish this mission. Therefore, i2CAT and a number of CIDAI partners have developed cutting-edge proofs of concept (POCs) applied to different sectors that validate the practical application of innovative IA and data technologies in areas of interest of the CATALONIA.AI strategy.
Deploying successful POCs is essential to a systematic AI adoption and to explore possible future applications. The POCs developed by CIDAI members are innovative solutions that apply to a variety of sectors such as healthcare, sustainability, industry 4.0 or agrifood.
“Light” Voice Assistant by i2CAT
With its POC, the i2CAT Foundation has strived to improve the privacy and security of language recognition systems, as well as response times and energy consumption. The “Light” Voice Assistant developed by i2CAT is a simple system that recognizes keywords and orders. It is capable of operating entirely locally, without depending at any time on an Internet connection or processing by external servers. The system is designed to work on home devices and uses an artificial intelligence algorithm that can be trained to detect custom keywords.
The objective of “Light” is to allow voice recognition in appliances with limited user interface and processing capabilities, such as refrigerators, washing machines or toasters. By integrating a local, low-cost processor into these appliances, they are able to recognize simple voice commands (a keyword followed by a command) without depending on personal voice assistants. This prevents a mandatory internet connection, an external server to process voice data, and the need to subscribe to the terms and conditions established by the voice assistant’s manufacturer.
Thanks to “Light”, users can incorporate voice recognition systems into their homes with the certainty that their data is safe and private and the system used is more robust than traditional voice assistants in the face of external internet attacks.