Detecção de Quedas e Crises Epilépticas com Dispositivo Vestível utilizando TinyML e ESP32
Resumo
A epilepsia é uma condição neurológica crônica que afeta cerca de 50 milhões de pessoas globalmente, demandando soluções que garantam a segurança e a agilidade no atendimento emergencial. Este trabalho apresenta o desenvolvimento de um sistema vestível baseado no microcontrolador ESP32 para o monitoramento contínuo de idosos e pacientes epilépticos. O objetivo central é a detecção autônoma de quedas e crises convulsivas por meio de sensores inerciais. A metodologia envolveu a criação de um conjunto de dados baseado em perfis de aceleração da literatura especializada, seguido pelo treinamento de um classificador de inteligência artificial na plataforma Edge Impulse. Utilizando a abordagem de TinyML, o modelo foi embarcado diretamente no hardware, permitindo inferências em tempo real com baixa latência e sem a necessidade do envio de dados a servidores, o que preserva a privacidade do usuário. Os resultados indicam que a computação de borda é eficaz para esta aplicação, permitindo o funcionamento do dispositivo de forma offline. Conclui-se que a solução oferece uma camada adicional de segurança, mitigando riscos de atraso no socorro e promovendo maior autonomia aos usuários.Referências
ARENDS, Johan BAM. Movement‐based seizure detection. Epilepsia, v. 59, p. 30-35, 2018.
ASADI-POOYA, Ali A. et al. Physical injuries in patients with epilepsy and their associated risk factors. Seizure, v. 21, n. 3, p. 165-168, 2012.
BENICZKY, Sándor et al. Seizure semiology: ILAE glossary of terms and their significance. Epileptic Disorders, v. 24, n. 3, p. 447-495, 2022.
ESPRESSIF SYSTEMS. ESP-NOW Wireless Communication Protocol, [S.d.]. Disponível em: [link]. Acesso em: 5 fev. 2026.
FÉRRER, H.; BARROS, H. Desenvolvimento de pulseira inteligente para detecção e prevenção de ataques epilépticos . Revista do Fórum Gerencial, v. 1, n. 2, p. 11, 20 out. 2021.
FISHER, Robert S. et al. Operational classification of seizure types by the International League Against Epilepsy: Position Paper of the ILAE Commission for Classification and Terminology. Epilepsia, v. 58, n. 4, p. 522-530, 2017.
FLEMING, Jane; BRAYNE, Carol. Inability to get up after falling, subsequent time on floor, and summoning help: prospective cohort study in people over 90. Bmj, v. 337, 2008.
GLASSTETTER, Martin et al. Identification of ictal tachycardia in focal motor-and nonmotor seizures by means of a wearable PPG sensor. Sensors, v. 21, n. 18, p. 6017, 2021.
GOODFELLOW, Ian et al. Deep learning. Cambridge: MIT press, 2016.
HYMEL, Shawn et al. Edge impulse: An mlops platform for tiny machine learning. arXiv preprint arXiv:2212.03332, 2022.
IOFFE, Sergey; SZEGEDY, Christian. Batch normalization: Accelerating deep network training by reducing internal covariate shift. In: International conference on machine learning. pmlr, 2015. p. 448-456.
JONES, Lliwen A.; THOMAS, Rhys H. Sudden death in epilepsy: Insights from the last 25 years. Seizure, v. 44, p. 232-236, 2017.
M5STACK. M5StickC: Documentation and Specifications. 2020. Disponível em: [link]. Acesso em: 3 fev. 2026.
NIJSEN, Tamara Mathea Elisabeth. Accelerometry based detection of epileptic seizures. 2008.
RAMGOPAL, Sriram et al. Seizure detection, seizure prediction, and closed-loop warning systems in epilepsy. Epilepsy & behavior, v. 37, p. 291-307, 2014.
RAY, Partha Pratim. A review on TinyML: State-of-the-art and prospects. Journal of King Saud University-Computer and Information Sciences, v. 34, n. 4, p. 1595-1623, 2022.
REUSCH, Rafael Schild; JURACY, Leonardo Rezende; MORAES, Fernando Gehm. Assessment and optimization of 1D CNN model for human activity recognition. In: 2022 XII Brazilian symposium on computing systems engineering (SBESC). IEEE, 2022. p. 17.
SEVAGJB. WT32-ETH01 + PEO Enclosure. Thingiverse, 4 nov. 2024. Disponível em: [link]. Acesso em: 23 fev. 2026.
WORLD HEALTH ORGANIZATION. Epilepsy. Genebra: WHO, 7 fev. 2024. Disponível em: [link]. Acesso em: 10 dez. 2025.
WORLD HEALTH ORGANIZATION. Step Safely: strategies for preventing and managing falls across the life-course. Geneva: World Health Organization, 2021. Disponível em: [link]. Acesso em: 15 jan. 2025.
ZIA, S. et al. Detection of Generalized Tonic Clonic Seizures and Falls in Unconstraint Environment Using Smartphone Accelerometer. IEEE Access, v. 9, p. 39432–39443, 2021.
ASADI-POOYA, Ali A. et al. Physical injuries in patients with epilepsy and their associated risk factors. Seizure, v. 21, n. 3, p. 165-168, 2012.
BENICZKY, Sándor et al. Seizure semiology: ILAE glossary of terms and their significance. Epileptic Disorders, v. 24, n. 3, p. 447-495, 2022.
ESPRESSIF SYSTEMS. ESP-NOW Wireless Communication Protocol, [S.d.]. Disponível em: [link]. Acesso em: 5 fev. 2026.
FÉRRER, H.; BARROS, H. Desenvolvimento de pulseira inteligente para detecção e prevenção de ataques epilépticos . Revista do Fórum Gerencial, v. 1, n. 2, p. 11, 20 out. 2021.
FISHER, Robert S. et al. Operational classification of seizure types by the International League Against Epilepsy: Position Paper of the ILAE Commission for Classification and Terminology. Epilepsia, v. 58, n. 4, p. 522-530, 2017.
FLEMING, Jane; BRAYNE, Carol. Inability to get up after falling, subsequent time on floor, and summoning help: prospective cohort study in people over 90. Bmj, v. 337, 2008.
GLASSTETTER, Martin et al. Identification of ictal tachycardia in focal motor-and nonmotor seizures by means of a wearable PPG sensor. Sensors, v. 21, n. 18, p. 6017, 2021.
GOODFELLOW, Ian et al. Deep learning. Cambridge: MIT press, 2016.
HYMEL, Shawn et al. Edge impulse: An mlops platform for tiny machine learning. arXiv preprint arXiv:2212.03332, 2022.
IOFFE, Sergey; SZEGEDY, Christian. Batch normalization: Accelerating deep network training by reducing internal covariate shift. In: International conference on machine learning. pmlr, 2015. p. 448-456.
JONES, Lliwen A.; THOMAS, Rhys H. Sudden death in epilepsy: Insights from the last 25 years. Seizure, v. 44, p. 232-236, 2017.
M5STACK. M5StickC: Documentation and Specifications. 2020. Disponível em: [link]. Acesso em: 3 fev. 2026.
NIJSEN, Tamara Mathea Elisabeth. Accelerometry based detection of epileptic seizures. 2008.
RAMGOPAL, Sriram et al. Seizure detection, seizure prediction, and closed-loop warning systems in epilepsy. Epilepsy & behavior, v. 37, p. 291-307, 2014.
RAY, Partha Pratim. A review on TinyML: State-of-the-art and prospects. Journal of King Saud University-Computer and Information Sciences, v. 34, n. 4, p. 1595-1623, 2022.
REUSCH, Rafael Schild; JURACY, Leonardo Rezende; MORAES, Fernando Gehm. Assessment and optimization of 1D CNN model for human activity recognition. In: 2022 XII Brazilian symposium on computing systems engineering (SBESC). IEEE, 2022. p. 17.
SEVAGJB. WT32-ETH01 + PEO Enclosure. Thingiverse, 4 nov. 2024. Disponível em: [link]. Acesso em: 23 fev. 2026.
WORLD HEALTH ORGANIZATION. Epilepsy. Genebra: WHO, 7 fev. 2024. Disponível em: [link]. Acesso em: 10 dez. 2025.
WORLD HEALTH ORGANIZATION. Step Safely: strategies for preventing and managing falls across the life-course. Geneva: World Health Organization, 2021. Disponível em: [link]. Acesso em: 15 jan. 2025.
ZIA, S. et al. Detection of Generalized Tonic Clonic Seizures and Falls in Unconstraint Environment Using Smartphone Accelerometer. IEEE Access, v. 9, p. 39432–39443, 2021.
Publicado
19/07/2026
Como Citar
QUINTO, Francisco Thiago Barbosa; LIMA, Thiago de Sena; BRITO, Fábio Timbó; SANTOS, José Daniel de Alencar.
Detecção de Quedas e Crises Epilépticas com Dispositivo Vestível utilizando TinyML e ESP32. In: WORKSHOP DE INOVAÇÃO, DESENVOLVIMENTO, EDUCAÇÃO E INCLUSÃO COM AÇÕES MAKER (IDEIA), 2. , 2026, Gramado/RS.
Anais [...].
Porto Alegre: Sociedade Brasileira de Computação,
2026
.
p. 62-71.
DOI: https://doi.org/10.5753/ideia.2026.21108.
