Depression is a mental illness that is difficult to diagnose and treat. This
mental disorder affects many older adults due several reasons, for instance
because of their physical limitations and the natural reduction of their social
circle. This article presents a system for monitoring the mood of the elderly
through voice processing. The system is particularly focused on detecting
sadness, which allows caregivers of family members to react on-time in
supporting the person in need. The sadness recognition is done by classifying
emotions in groups, according to the Circumflex Model of Affect. After
evaluating the system using several emotion databases, the obtained results
indicate that this solution is able to recognize 94% of the cases in men and
79% in women. This solution can be embedded in ubiquitous systems that monitor
the mood of people in several scenarios.