Study On Calorie Consumption Based On Human Activity Recognition Technology

Automatic identification of physical activities performed by humans is known as Human Activity Recognition(HAR).There are several techniques to measure these characteristics of movement during physical activities,such as Inertial Measurement Units(IMU).IMUs are fundamental in this context and are characterized by their flexibility of use,low cost and less impact on privacy.With the use of Notch sensors,it is possible to sample some measurements,such as the acceleration and angular velocity of a body,and use this information to learn models that are capable of correctly classifying activities in their corresponding classes.In this article,we propose to use Convolutional Neural Networks(CNN)to classify human activities.Our models use raw data obtained through Notch sensors.We explored various activities,showing how motion signals can be adapted to feed them to CNN through the use of different network architectures.The experimental results obtained in a data set of 6 lower and upper limb activities,collected from a group of participants with the use of eleven different sensors,are very promising.