Purpose: Data provided by the World Health Organization indicate that the number of people aged 60 years or more in the world population is gradually increasing. Moreover, projections clarify that in 2030 there will be 1.4 billion elderly people and 2.1 billion in 2050. Along with the natural aging process of any human being, changes in perception and cognition can cause damage that contributes to elderly people having difficulties both to recognize and to express emotions through the face. Few researches in the literature address the recognition of emotions in elderly, whether they are affected by dementia processes or not. In consequence of this, few databases are made available for carrying out work and experiments. Not being able to express and recognize emotions through the face can contribute to the elderly having difficulties in communicating important messages, which may even compromise their physical integrity.Methods: Therefore, this work aims to develop an application for Emotion Recognition in the elderly through Facial Expressions. For this, we first used Haar cascade Frontal Face for face detection and implemented a Convolutional Neural Network to classify emotions, using FER2013 database for training, validation and testing. In a second part of the methodology, in order to assess the performance of the algorithm in this context, we applied the developed model to recognize emotions in static images of elderly people.Results: As a result, the accuracy achieved by the developed model was 0.6375. From the images tested, for 52.63% of them the model was able to detect the face and identify some emotion. On the other hand, in 47.37% of the images, the model had difficulty both in detecting the face and in identifying emotions.Conclusion: Finally, the findings and discussions exposed in this work are promising, we also found and shared limitations and related them to our goals for future works. The possibility of developing intelligent systems that support emotion recognition in elderly population emerges as a valuable tool, representing an alternative to promote not only quality of life for the elderly themselves, but also for the entire support network around them.

No abstract

No abstract

No abstract

Ambientes hospitalares precisam de refrigeradores hospitalares para armazenar fármacos, vacinas, bolsas de sangue, dentre outros. Tais equipamentos são configurados de forma a manter determinada faixa de temperatura, visto que os produtos armazenados são sensíveis a mudanças de temperatura fora dessa faixa. Este projeto objetiva analisar as variações de temperatura acima do adequado. Nos experimentos realizados foram implementados diferentes técnicas de detecção de anomalias utilizando três métodos de agrupamento: k-means, DBSCAN e Isolation Forest. Levando em consideração a acurácia encontrada (76,7%), o método utilizado foi o DBSCAN. Com a análise realizada, foi possível perceber diversas relações entre os valores de temperatura, quantidade de alertas e os horários que eles aconteceram. Observou-se que a maior parte das anomalias encontradas aconteceram entre às 6:00 e às 8:00 horas da manhã, coincidindo com o horário de troca de turnos entre funcionários.