Speech recognition based home automation application

Home automation applications have attracted a lot of interest of researchers and companies for many years. Home owners are usually using tablets, smart phone applications or even control panels to act upon the household appliances, i.e. for turning on/off the lights, the washing machine, moving up/down the blinds, etc.

Lately, speech recognition has also made its way into home automation applications, using devices such as Alexa, but also smart phone applications for Siri, Google, etc. However, if the speech-recognition based systems are not properly designed to be secure, one can consider that they can become the Pandora’s box for intruders.

One such example is described here [1] where a person had spent a lot of money to make his house “smart”, but a neighbour just by yelling at Siri to “open the front door” from outside the household, he was able to easily get in.

To address such issues, RERUM prepared a joint use case with the EU-H2020-MSCA-RISE project LISTEN for secure speech-recognition based home automation applications. In this example, we have a microphone array (built within LISTEN) with 8 digital microphones, able to calculate the direction of arrival of the voice signal. Then, RERUM assigns different roles to the “users” according to the direction of the speech signal and identifies if the users are authorized to act upon the household appliances or not. Assuming that the microphone array can be placed either at the front door or at the windows, the system can easily distinguish if a user is inside or outside the house, so it can reject any commands coming from outside (so in the example of the neighbour, the front door wouldn’t open).

See below a video demonstrating the integration between RERUM and LISTEN and the implementation of the above mentioned scenario.

[1] https://nakedsecurity.sophos.com/2016/09/22/siri-opens-smart-lock-to-let-neighbor-walk-into-a-locked-house/


Bookmark the permalink.