Single Triaxial Accelerometer-Gyroscope Classification for Human Activity Recognition

Abstract

Evaluated activity as a detail of the human physical movement has become a leading subject for researchers. Activity recognition application is utilized in several areas, such as living, health, game, medical, rehabilitation, and other smart home system applications. For recognizing the activity, the accelerometer was popular sensors. As well as a gyroscope, in addition to dimension, low computation, and can be embedded in a smartphone. Used smartphone with an accelerometer as a popular solution for recognized daily activity. Signal was generated from the accelerometer as a time-series data is an actual approach like a human activity pattern. Traditional machine learning method in mid of the modern method worth it considering. Single position triaxial accelerometer-gyroscope Motion data have acquired in an of 30 volunteers.

Human Behavior Recognition Model Based on Feature and Classifier Selection

“…The constructed feature vector was also trained and tested on other classifier packages—DT, RF, KNN, LR and ECLF classifiers. The comparison among the recognition rate obtained and those reported in reference [ 32 ] is shown in Table 3 .…”

“…To identify basic actions with different dynamic characteristics, it is necessary to extract a large number of features, which not only causes information redundancy, but also brings a large amount of calculation to the computer, making the time of model training relatively long. In the experimental method (method 3) in this section, 126 frequency-domain features proposed in this paper were selected to classify 6 types of actions and the effect of the SVM classifier was improved compared with the literature [ 32 ]. The constructed feature vector was also trained and tested on other classifier packages—DT, RF, KNN, LR and ECLF classifiers.…”

“…The authors of [ 53 ] used 248 features and the accuracy of several depth learning methods was lower than that of traditional methods. This study selected the same experimental setting as that in the literature [ 32 ]. The authors of [ 32 ] did not carry out feature selection and directly used 561 time-domain and frequency-domain features extracted during data acquisition.…”

“…Bhuiyan et al [ 31 ] used a multi-class support vector machine (MC-SVM) as a classifier to classify five kinds of daily actions performed by humans and achieved good recognition performance. Minarno [ 32 ] compared the classification performance of DT, RF, KNN, logistic regression (LR), SVM and ensemble voting classifier (ECLF). The accuracy of the logistic regression method was 98% and the accuracy of the support vector machine classification method was 93.86%.…”

Multiscale Bidirectional Input Convolutional and Deep Neural Network for Human Activity Recognition

“…Due to its small size and convenient portability, the built-in sensors are becoming more and more diverse, and specific types of activities can be effectively classified through the information of multiple sensors. For example, the built-in accelerometers of smart phones [10,11] can describe human actions, such as standing, walking, and running [12,13]. Similarly, by collecting audio information from the phone microphone [14], the user's activities can be identified, such as listening to music, speaking, and sleeping [15,16], running rhythm can be monitored [17], and user respiratory symptoms are related to sound, such as sneezing or coughing [15].…”