The Social Robotics Group of the Robotics Lab at the Universidad Carlos III de Madrid (UC3M) has developed an innovative methodology based on artificial intelligence (AI) that allows a pet-type assistive robot to recognize the people it interacts with. The prototype, named Mía, is currently being tested in Madrid City Council day centers for the personalized affective stimulation of elderly people with cognitive decline.
One of the greatest challenges in current social robotics is getting machines to interact with us in a natural and personalized way. To achieve this, one approach is for the robot to know who it is talking to. The new user voice recognition system developed by this UC3M scientific team, recently published in the journal Applied Sciences, operates "onboard" Mía, a robotic rabbit that simulates a pet.
"We use this robot for affective stimulation for the elderly with cognitive decline, and user recognition is an important step towards the robot having a personalized behavior adapted to the needs of each individual," explains one of the authors of this work, José Carlos Castillo Montoya, from the Social Robotics Group of the Robotics Lab at UC3M.
Unlike other traditional recognition systems that rely on cameras and image processing (which usually require high computing capacities and can violate privacy), this proposal exclusively uses the robot's built-in microphone. Furthermore, it operates locally and incrementally, which offers two advantages according to the researchers. On one hand, greater privacy, as the voiceprints never leave the physical device itself. On the other hand, a sort of on-the-fly learning:
"The robot does not need a setup phase to learn your voice; it hears you speak for the first time and automatically generates a profile," notes another author of the study, Arecia Segura Bencomo, from UC3M's Department of Systems Engineering and Automation.
Benefits of animal robotics and affective stimulation
This advancement implemented in Mía, a rabbit-shaped robotic prototype developed entirely at UC3M, falls within the scope of animal robotics. This is a therapeutic paradigm designed to transfer the proven benefits of therapy with real animals (such as reducing stress and anxiety) to controlled environments where animal welfare or the patient's capacity for care limit their use.
"In the specific case of the elderly with cognitive decline, the robot activates the need to care for something," points out Professor Castillo. "We have verified in pilot tests at Madrid City Council day centers that, in addition to improving the elderly person's mood, the robot acts as a social catalyst: it breaks isolation, mitigates unwanted loneliness, and encourages users to start interacting more with each other."
An artificial intelligence-based "voice signature"
The system's operation emulates human learning capacity. Every time a person speaks to the robot, the algorithm analyzes the nuances and patterns of the acoustic signal to create a unique "voice signature".
"Our proposal is simple because the robot's computing capacities are very limited," explains José Carlos Castillo Montoya. "The algorithm generates a sort of map and, using open-source code that we have improved in our laboratory, groups together signatures that resemble each other. Ultimately, each group corresponds to a specific user."
The system stands out especially for its ability to deal with unforeseen situations in real time. If it interacts with an unknown person, the AI detects that their signature does not fit into any previous group and, if the interaction continues, it is capable of generating a completely new dynamic profile, learning that this is a new user. In experimental tests, the software achieved a high success rate in identifying known users.
This advancement opens the door to a new generation of much more accessible assistive and companion robots, capable of entering a home and autonomously learning who its members are just by hearing the first "good morning". The research team highlights the future potential for personalization that this technology unlocks.
"We want the robot to be able to adapt by reacting differently depending on who is in front of it. If it detects a specific user who tends to be more restless, specific mechanisms and behaviors aimed at calming that person will be automatically activated."
Currently, the research team has several operational prototypes of Mía with which they continue to mature the technology in real clinical scenarios, while seeking alliances with companies interested in transferring this knowledge and bringing these assistive platforms to the general market.
Video: https://youtu.be/1vUJdQiaDlA