SIRAH: sistema de reconhecimento de atividades humanas e avaliação do equilibrio postural
Carregando...
Data
2017-06-09
Autores
Durango, Melisa de Jesus Barrera [UNESP]
Título da Revista
ISSN da Revista
Título de Volume
Editor
Universidade Estadual Paulista (Unesp)
Resumo
O reconhecimento de atividades humanas abrange diversas técnicas de classificação que permitem identificar padrões específicos do comportamento humano no momento da ocorrência. A identificação é realizada analisando dados gerados por diversos sensores corporais, entre os quais destaca-se o acelerômetro, pois responde tanto à frequência como à intensidade dos movimentos. A identificação de atividades é uma área bastante explorada. Porém, existem desafios que necessitam ser superados, podendo-se mencionar a necessidade de sistemas leves, de fácil uso e aceitação por parte dos usuários e que cumpram com requerimentos de consumo de energia e de processamento de grandes quantidades de dados. Neste trabalho apresenta-se o desenvolvimento do Sistema de Reconhecimento de atividades Humanas e Avaliação do Equilíbrio Postural, denominado SIRAH. O sistema está baseado no uso de um acelerômetro localizado na cintura do usuário. As duas fases do reconhecimento de atividades são apresentadas, fase Offline e fase Online. A fase Offline trata do treinamento de uma rede neural artificial do tipo perceptron de três camadas. No treinamento foram avaliados três estudos de caso com conjuntos de atributos diferentes, visando medir o desempenho do classificador na diferenciação de 3 posturas e 4 atividades. No primeiro caso o treinamento foi realizado com 15 atributos, gerados no domínio do tempo, com os que a rede neural artificial alcançou uma precisão de 94,40%. No segundo caso foram gerados 34 atributos tanto no domínio do tempo quanto da frequência. Uma vez treinada a rede neural artificial, esta classificou corretamente as posturas e atividades em 94,35% das vezes. No terceiro caso aplicou-se o método de seleção automática PCA (Principal Componente Analysis) no conjunto de 34 atributos. Com 20 componentes principais a rede neural artificial teve uma precisão de 95,55%. Na fase Online foi implementado no sistema SIRAH o modelo obtido no treinamento com o primeiro estudo de caso, com o objetivo de classificar dados novos do sensor em tempo real. Estes atributos são gerados exclusivamente no domínio do tempo, evitando a implementação da Transformada de Fourier no sistema, que aumentaria a complexidade computacional e o tempo de execução do algoritmo. O sistema SIRAH envia ao computador a atividade ou postura classificada, limitando desta forma a quantidade de informação transmitida na rede local. O sistema SIRAH também calcula a raiz do valor quadrático médio da aceleração horizontal e os deslocamentos nos eixos ântero-posterior e médio-lateral do centro de massa do corpo, indicadores da oscilação corporal, parâmetros sugeridos como medida do equilíbrio na postura ereta semi-estática. Para validar esta funcionalidade foram realizados testes que alteram o sistema de controle postural, usados para avaliar o equilíbrio. Os resultados dos testes mostraram que o acelerômetro consegue diferenciar estas condições, concluindo-se que esta funcionalidade torna ao sistema SIRAH uma ferramenta válida na avaliação da oscilação corporal continua, que pode ser usada em ambientes externos a clínicas e laboratórios especializados. SIRAH, com suas duas funcionalidades, é um sistema confortável e de fácil uso, pois sua localização não atrapalha os movimentos dos usuários.
Human activity recognition encompasses varying classification techniques that identify of specific patterns of human behavior at the time of their occurrence. Identification is achieved by analyzing data generated by several body sensors, in particular the accelerometer, as it responds to both the frequency and the intensity of the movements. Human activity recognition is a well-explored area. However, there are challenges that need to be overcome, such as the need for light systems that are easy to use, user-friendly, that meet energy consumption requirements and can processes large amounts of data. In this research is presented the development of the of Human Activity Recognition and Balance Assessment System, denominated SIRAH. The system is based on an accelerometer used in the user's waist. Two phases of activity recognition are presented, offline phase and online phase. The offline phase deals with the training of an Artificial Neural Network. In the training, three case studies with different sets of features were evaluated in order to measure the performance of the classifier on the classification of 3 postures and 4 activities. In the first case, the Artificial Neural Network achieved an accuracy of 94.40% using 15 features generated in the time domain. In the second case, 34 features were generated in both time and frequency domains. Once the Artificial Neural Network was trained, it correctly classified the postures and activities in 94.35% of the instances. In the third case, the PCA (Principal Component Analysis) automatic selection method was applied to the second case study. With 20 components, the Artificial Neural Network had an accuracy of 95.55%. In the online phase, the model obtained training for the first case study was implemented in the SIRAH system to classify new sensor data in real time. These attributes are generated exclusively in the time domain and so the need to implement the Fourier Transform in the system is avoided. This would increase the computational complexity and the execution time of the algorithm. The SIRAH system only sends the classified activity or posture to the computer, thereby limiting the amount of information transmitted in the local network. The SIRAH system also calculates root mean square value for horizontal accelerations and the displacements on the anteroposterior and mediolateral axes of the body mass center, body sway indicators, suggested parameters for body sway measurement in quiet standing. To validate this functionality were performed tests that alter the balance control system, used to evaluate a person's balance. The results of the tests showed that the accelerometer can differentiate these conditions, concluding that the SIRAH system is a valid tool in the assessment of continuous body sway, which can be used in environments outside specialized clinics and laboratories. SIRAH, with its two functionalities, is a comfortable and easy to use system, because its location does not disturb the movements of the users.
Human activity recognition encompasses varying classification techniques that identify of specific patterns of human behavior at the time of their occurrence. Identification is achieved by analyzing data generated by several body sensors, in particular the accelerometer, as it responds to both the frequency and the intensity of the movements. Human activity recognition is a well-explored area. However, there are challenges that need to be overcome, such as the need for light systems that are easy to use, user-friendly, that meet energy consumption requirements and can processes large amounts of data. In this research is presented the development of the of Human Activity Recognition and Balance Assessment System, denominated SIRAH. The system is based on an accelerometer used in the user's waist. Two phases of activity recognition are presented, offline phase and online phase. The offline phase deals with the training of an Artificial Neural Network. In the training, three case studies with different sets of features were evaluated in order to measure the performance of the classifier on the classification of 3 postures and 4 activities. In the first case, the Artificial Neural Network achieved an accuracy of 94.40% using 15 features generated in the time domain. In the second case, 34 features were generated in both time and frequency domains. Once the Artificial Neural Network was trained, it correctly classified the postures and activities in 94.35% of the instances. In the third case, the PCA (Principal Component Analysis) automatic selection method was applied to the second case study. With 20 components, the Artificial Neural Network had an accuracy of 95.55%. In the online phase, the model obtained training for the first case study was implemented in the SIRAH system to classify new sensor data in real time. These attributes are generated exclusively in the time domain and so the need to implement the Fourier Transform in the system is avoided. This would increase the computational complexity and the execution time of the algorithm. The SIRAH system only sends the classified activity or posture to the computer, thereby limiting the amount of information transmitted in the local network. The SIRAH system also calculates root mean square value for horizontal accelerations and the displacements on the anteroposterior and mediolateral axes of the body mass center, body sway indicators, suggested parameters for body sway measurement in quiet standing. To validate this functionality were performed tests that alter the balance control system, used to evaluate a person's balance. The results of the tests showed that the accelerometer can differentiate these conditions, concluding that the SIRAH system is a valid tool in the assessment of continuous body sway, which can be used in environments outside specialized clinics and laboratories. SIRAH, with its two functionalities, is a comfortable and easy to use system, because its location does not disturb the movements of the users.
Descrição
Palavras-chave
Inteligência ambiental, Ambientes inteligentes, Internet das coisas, Reconhecimento de atividades humanas, Sensores, Sensores corporais, Acelerômetro, Classificação em tempo real, Redes neurais artificiais, Oscilação corporal, Baropodômetro, Ambient intelligence: smart environments, Internet of things, Human activity recognition, Sensors, Wearable sensor, Accelerometer, Online classification, Real-time classification, Artificial neural network, Principal component analysis, Body sway, Baraopodometer