Análise do posicionamento e deslocamento articular em exercícios resistidos por modelos ocultos de Markov: desenvolvimento e validação do método
Carregando...
Arquivos
Data
2020-09-11
Autores
Peres, André Bairros
Título da Revista
ISSN da Revista
Título de Volume
Editor
Universidade Estadual Paulista (Unesp)
Resumo
Identificar a estabilidade do posicionamento articular humano permite a análise de perfis comuns durante as ações motoras de transporte de objetos e do próprio corpo. Esta análise apresenta grande relevância em algumas atividades profissionais que exigem a comparação entre seres, objetos ou ações. Dentre as atividades laborais nas quais o reconhecimento de padrões se torna importante está a postura, o desempenho motor e o distúrbio locomotor patológico. O uso de tecnologias permite ao profissional obter informações mais precisas durante a análise dos diferentes perfis motores e permitir um julgamento apropriado quanto à necessidade de melhorias de controle das ações. Os modelos ocultos de Markov - MOM são modelos estatísticos com grande aplicabilidade e eficiência no reconhecimento de padrões. Este trabalho analisou o uso de MOM no reconhecimento de alterações de posições articulares que modificam a execução de referência de exercícios físicos resistidos: rosca direta e agachamento livre. Dez voluntários realizaram três execuções de cada exercício com diferentes cargas: 0%, 25% e 50% do peso corporal para o exercício de rosca e 0%, 50% e 75% para o agachamento. As execuções foram gravadas em câmera de vídeo digital, considerando apenas movimentos no plano sagital. Dados de deslocamentos de articulações foram obtidos mediante captura de coordenadas cartesianas das trajetórias percorridas. Análises de variância identificaram as execuções que sofreram alterações significativas (p < 0,05) de valores de deslocamento com as alterações de carga. MOM treinados com dados de execuções sem carga conseguiram distinguir as execuções que sofreram alterações do padrão de movimentos pelo aumento da carga. Profissionais da área também analisaram os movimentos das articulações e os resultados de suas avaliações atingiram até 90% de concordância com o modelo proposto. Os resultados obtidos mostraram que MOM apresentam viabilidade na identificação de alterações de padrão de movimentos mono e multiarticulares, com sobrecarga, pela análise de posicionamento articular. Mostraram também possibilidade de serem avaliados os movimentos registrados em vídeos obtidos sem qualquer referência métrica de dados. Com a participação de avaliadores humanos também se mostrou o alinhamento do método com a interpretação profissional de execuções adequadas.
Identifying the stability of the human joint position allows the analysis of common profiles during the motor actions of transporting objects and the body itself. This analysis is highly relevant in some professional activities that require comparison between beings, objects, or actions. Among the work activities in which pattern recognition becomes important are posture, motor performance, and pathological disturbance of locomotor system. The use of technologies allows the professional to obtain information that is more accurate during the analysis of the different locomotor profiles and allows an appropriate judgment as to the need for control improvements. The hidden Markov models - HMM are statistical models with great applicability and efficiency in pattern recognition. This work analyzed the use of HMM in recognizing changes in joint positions that modify the reference execution of resisted physical exercises: barbell curls and free squats. Ten volunteers performed three executions of each exercise with different loads: 0%, 25%, and 50% of the bodyweight for the biceps exercise and 0%, 50%, and 75% for the squat exercise. The executions were recorded on a digital video camera, considering only movements in the sagittal plane. Joint displacement data were obtained by capturing Cartesian coordinates of the trajectories covered. Analysis of variance identified the executions that underwent significant changes (p = 0.05) of displacement values with changes in load. HMM trained with data of executions without load were able to distinguish the executions that suffered changes in the movement pattern due to the increase in the load. Professionals in the area also analyzed the movements of the joints and the results of their assessments reached up to 90% agreement with the proposed model. The results obtained showed that MOM is feasible in identifying changes in the pattern of mono and multiarticular movements, with overload, by analyzing joint positioning. They also showed the possibility of evaluating movements recorded in videos obtained without any metric data reference. With the participation of human evaluators, the alignment of the method with the professional interpretation of appropriate executions was also shown.
Identifying the stability of the human joint position allows the analysis of common profiles during the motor actions of transporting objects and the body itself. This analysis is highly relevant in some professional activities that require comparison between beings, objects, or actions. Among the work activities in which pattern recognition becomes important are posture, motor performance, and pathological disturbance of locomotor system. The use of technologies allows the professional to obtain information that is more accurate during the analysis of the different locomotor profiles and allows an appropriate judgment as to the need for control improvements. The hidden Markov models - HMM are statistical models with great applicability and efficiency in pattern recognition. This work analyzed the use of HMM in recognizing changes in joint positions that modify the reference execution of resisted physical exercises: barbell curls and free squats. Ten volunteers performed three executions of each exercise with different loads: 0%, 25%, and 50% of the bodyweight for the biceps exercise and 0%, 50%, and 75% for the squat exercise. The executions were recorded on a digital video camera, considering only movements in the sagittal plane. Joint displacement data were obtained by capturing Cartesian coordinates of the trajectories covered. Analysis of variance identified the executions that underwent significant changes (p = 0.05) of displacement values with changes in load. HMM trained with data of executions without load were able to distinguish the executions that suffered changes in the movement pattern due to the increase in the load. Professionals in the area also analyzed the movements of the joints and the results of their assessments reached up to 90% agreement with the proposed model. The results obtained showed that MOM is feasible in identifying changes in the pattern of mono and multiarticular movements, with overload, by analyzing joint positioning. They also showed the possibility of evaluating movements recorded in videos obtained without any metric data reference. With the participation of human evaluators, the alignment of the method with the professional interpretation of appropriate executions was also shown.
Descrição
Palavras-chave
Reconhecimento de padrões, Movimento humano, Rosca direta, Agachamento livre, Pattern recognition, Human movement, Biceps curl, Free squat