A robot that knows when an older person is feeling sad

‘In order to support caregivers, we want the Guardian care robot to be able to detect the feelings of an older person’, says Alexandre de Masi, PhD Student of the University of Geneva who is involved in the research. ‘With Artificial Intelligence (AI) the robot will improve the interpretation of emotions over time.’

His grandparents will turn ninety in a couple of years and are a strong motivation for his work. De Masi: ‘It would be wonderful if they can live in their home as long as possible. As involved researchers, we all have to ask ourselves this question: “How can I help my parents or grandparents in the future?”

Maybe you should check on your loved one?

The support of a care robot such as Guardian could help older people to live longer at home. Emotional recognition plays an important role in the developing process. De Masi: ‘This emotional recognition will happen through audio signals or an image or video of the face. When the older person is sad, the robot can inform the informal caregiver about this through an app or other device. For instance, with the message: “Maybe you should check on your loved one to hear if he’s okay.”


Distinguishing sounds

De Masi: ‘Right now, we are focusing only on audio and video features. We are also working on presence detection with the University of Politecnica delle Marche (Italy). The first step is to limit the data acquisition. Not every sound says something about the older person. The robot has to exclude sounds that are coming from a television, radio, or a different person in the room. With training data we can also recognize emotions from the way someone speaks. We are conducting a sentiment analysis when someone is speaking to the robot. When someone expresses joy with a laugh, it’s easy. But with other emotions, it can be more difficult.’

Cultural bias

In his work, he encounters other challenges as well. De Masi: ‘We are using publicly available training data, but it could be that there is some cultural bias involved. The emotion recognized by the software is not necessarily the emotion that the person is feeling inside. We are developing this software for older people in Europe, but in Switzerland a feeling could be expressed differently than in Italy. That’s why correction could be necessary in the testing phase in different European countries.’

Co-creation is important

‘We are also exploring how to present the information to the informal and formal caregiver. For example, are we are using a timeline, emojis or notifications? We don’t have all the answers right now. That is why the co-creation process with caregivers and older persons in the near future is so important.’

What about privacy?

Extracting information from audio and video recordings can also bring up ethical and privacy issues. De Masi: ‘We have to be very careful with the data. We are aiming to process the data as much as possible on the robot itself, so that most of the information will not be sent to an external server. And we want to work only with cloud parties that have a license to store medical information and are compliant with the General Data Protection Regulation (GDPR).’

Making the knowledge available

De Masi is glad that the University of Geneva is involved in the Guardian project. ‘Ageing will bring challenges in Switzerland as well. Technology can be one of the answers to the increasing care demand and we can take important steps in this research. In regard to emotional recognition, not a lot of research has been conducted with older people. My supervisor is professor Kararzyna Wac and she is engaged in the standardisation of the work that we are doing for Guardian. We want to measure our performance in this project to make the knowledge available for other parties as well. This could be in the form of an article in a scientific paper, for example.’