Pattern recognition techniques for the identification of Activities of Daily Living using mobile device accelerometer

Abstract

This paper focuses on the recognition of Activities of Daily Living (ADL) applying pattern recognition techniques to the data acquired by the accelerometer available in the mobile devices. The recognition of ADL is composed by several stages, including data acquisition, data processing, and artificial intelligence methods. The artificial intelligence methods used are related to pattern recognition, and this study focuses on the use of Artificial Neural Networks (ANN). The data processing includes data cleaning, and the feature extraction techniques to define the inputs for the ANN. Due to the low processing power and memory of the mobile devices, they should be mainly used to acquire the data, applying an ANN previously trained for the identification of the ADL. The main purpose of this paper is to present a new method implemented with ANN for the identification of a defined set of ADL with a reliable accuracy. This paper also presents a comparison of different types of ANN in order to choose the type for the implementation of the final method. Results of this research probes that the best accuracies are achieved with Deep Learning techniques with an accuracy higher than 80%. Following the previous research studies [4][5][6], the recognition of the ADL is composed by several steps, such as the data acquisition, the data processing, composed by data cleaning, data imputation, and feature extraction, the data fusion, and the artificial intelligence methods for the concrete identification of the ADL. However, this study only uses the accelerometer sensor, removing some steps of the proposed architecture. Based on the assumption that the sensor was always acquiring the data, the final steps used are the data acquisition, the data cleaning, and the application of the artificial intelligence methods.During the last years, the recognition of ADL has been studies by several authors [7][8][9][10][11][12], verifying that the Artificial Neural Networks (ANN) are widely used. This paper proposes the creation of a method for the recognition of ADL using accelerometer, comparing three types of ANN, such as Multilayer Perception (MLP) with Backpropagation, Feedforward neural network with Backpropagation, and Deep Learning, in order to verify the method that achieves the best accuracy in the recognition of running, walking, going upstairs, going downstairs, and standing. The datasets are composed with raw accelerometer data acquired by individual with ages ranged between 16 and 60 years old and different lifestyles, with a mobile device in the pocket. For the implementation of these types of ANN are used several datasets with different sets of features, identifying the best features to increase the accuracy of the recognition, and three Java libraries are used for the implementation of the different methods, such as Neuroph [13], Encog [14], and DeepLearning4j [15], achieving the best accuracy with Deep Learning methods.The remaining sections of this paper are organized as follows: Section 2 presents a brief literature review re...

Data Fusion on Motion and Magnetic Sensors embedded on Mobile Devices for the Identification of Activities of Daily Living

“…Due to the limitations of mobile devices and regarding the results obtained with the method for the recognition of ADL with the accelerometer previously performed, presented in [13], which was verified that the best results were obtained with Deep Neural Networks with L2 regularization and normalized data with an accuracy of 85.89%.…”

“…In the figure 7, a simplified schema for the development of a framework for the identification of ADL is presented. According to the previous study based only in the use of the accelerometer sensor for the recognition of ADL, presented in [13], the best results achieved for each type of neural network are presented in the table 6, verifying that the best method is Deep Neural Networks with normalized data, reporting an accuracy of 85.89%. In the case of the mobile device only has the accelerometer sensor available, Deep Neural Networks with normalized data should be implemented in the framework for the recognition of ADL, removing the data fusion, as presented in the figure 7.…”

“…After this research, the methods that should be implemented in the framework for the recognition of ADL defined in [4][5][6] is a function of the number of sensors available in the off-the-shelf mobile device. According to the results available in [13], if the mobile device has only the accelerometer sensor, the method that should be implemented is the Deep Neural Networks, verifying with this research the best methods for the use of the datasets defined in the sections 3.3.1 and 3.3.2.…”

“…Extending a previous study [13] that used only the accelerometer sensor, this study fuses the features extracted from the accelerometer and magnetometer sensors (section 3.3.1), and the features extracted from the accelerometer, gyroscope and magnetometer sensors (section 3.3.2). Finally, the artificial intelligence methods for the identification of ADL are presented in the section 3.3.3.…”

User Environment Detection with Acoustic Sensors Embedded on Mobile Devices for the Recognition of Activities of Daily Living

“…The advantages of recognition of the environments are not limited to the increasing of the number of ADL recognized, but it allows the framework to combine the environments with the ADL recognition returning different results, e.g., the user is walking on the street.The topic related to the recognition of the ADL has some studies available in the literature [8][9][10][11][12][13], but there are no studies that uses all sensors available on the off-the-shelf mobile devices, however the Artificial Neural Networks (ANN) is one of the most used methods in this topic. Based on our previous studies using motion and magnetic sensors for the development of the framework for the recognition of ADL and their environments [4,14], this study proposes the creation of several methods to adapt the framework to the number of sensors available in off-the-shelf mobile devices. Some methods using different combinations of sensors are presented in previous studies [4,14], such as the method using accelerometer data, the method using accelerometer and magnetometer data, and the method using accelerometer, magnetometer and gyroscope sensors.…”