LONG-REMI: An AI-Based Technological Application to Promote Healthy Mental Longevity Grounded in Reminiscence Therapy

Abstract

Reminiscence therapy (RT) consists of thinking about one's own experiences through the presentation of memory-facilitating stimuli, and it has as its fundamental axis the activation of emotions. An innovative way of offering RT involves the use of technology-assisted applications, which must also satisfy the needs of the user. This study aimed to develop an AI-based computer application that recreates RT in a personalized way, meeting the characteristics of RT guided by a therapist or a caregiver. The material guiding RT focuses on intangible cultural heritage. The application incorporates facial expression analysis and reinforcement learning techniques, with the aim of identifying the user's emotions and, with them, guiding the computer system that emulates RT dynamically and in real time. A pilot study was carried out at five senior centers in Barcelona and Portugal. The results obtained are very positive, showing high user satisfaction. Moreover, the results indicate that the high frequency of positive emotions increased in the participants at the end of the intervention, while the low frequencies of negative emotions were maintained at the end of the intervention.

Keywords: cognitive impairment; emotions recognition; face tracking; intangible cultural heritage; reinforcement learning; reminiscence therapy.

Figure 1

Example of identification of a face with the corresponding smile index. The smile index in this case is of 0.99, meaning that the person is clearly smiling.

Figure 2

Example of screenshots of the application (in Catalan). The application contains material in Catalan and Portuguese as it is intended for citizens of both countries.

Figure 3

Volunteers during a LONG-REMI session.