ADL Classification Based on Autocorrelation Function of Inertial Signals

Gomaa, Walid; Elbasiony, Reda; Ashry, Sara

doi:10.1109/icmla.2017.00-53

Cited by 16 publications

(7 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Moufawad et al [19] used several features in a DTs algorithm to differentiate nine classes of ADL, and an overall accuracy of 97% was obtained across all activities. An accuracy of 80% for a set of 14 ADL classified was attained by Gomaa et al [4], and the smallest average sensitivity and specificity achieved by the proposed Random Forest (RF) classification algorithm were 81% and 98%, respectively. Two ML algorithms were implemented by Gupta et al [20] for the classification of six classes of ADL-Naive Bayes (NB) and K-NN.…”

Section: Introductionmentioning

confidence: 99%

“…Several activities are being recognized nowadays. ADL related to human locomotion, such as walking, running, moving up and down stairs, or just sitting or lying down, are identified in several papers [2][3][4][5][6]. Other activities involving finer gestures with the upper limbs, such as driving, talking on the phone, or eating are also addressed [3,7].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Inertial Data-Based AI Approaches for ADL and Fall Recognition

Martins

Ribeiro

Soares

et al. 2022

Sensors

View full text Add to dashboard Cite

The recognition of Activities of Daily Living (ADL) has been a widely debated topic, with applications in a vast range of fields. ADL recognition can be accomplished by processing data from wearable sensors, specially located at the lower trunk, which appears to be a suitable option in uncontrolled environments. Several authors have addressed ADL recognition using Artificial Intelligence (AI)-based algorithms, obtaining encouraging results. However, the number of ADL recognized by these algorithms is still limited, rarely focusing on transitional activities, and without addressing falls. Furthermore, the small amount of data used and the lack of information regarding validation processes are other drawbacks found in the literature. To overcome these drawbacks, a total of nine public and private datasets were merged in order to gather a large amount of data to improve the robustness of several ADL recognition algorithms. Furthermore, an AI-based framework was developed in this manuscript to perform a comparative analysis of several ADL Machine Learning (ML)-based classifiers. Feature selection algorithms were used to extract only the relevant features from the dataset’s lower trunk inertial data. For the recognition of 20 different ADL and falls, results have shown that the best performance was obtained with the K-NN classifier with the first 85 features ranked by Relief-F (98.22% accuracy). However, Ensemble Learning classifier with the first 65 features ranked by Principal Component Analysis (PCA) presented 96.53% overall accuracy while maintaining a lower classification time per window (0.039 ms), showing a higher potential for its usage in real-time scenarios in the future. Deep Learning algorithms were also tested. Despite its outcomes not being as good as in the prior procedure, their potential was also demonstrated (overall accuracy of 92.55% for Bidirectional Long Short-Term Memory (LSTM) Neural Network), indicating that they could be a valid option in the future.

show abstract

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Inertial Data-Based AI Approaches for ADL and Fall Recognition

Martins

Ribeiro

Soares

et al. 2022

Sensors

View full text Add to dashboard Cite

show abstract

Section: Motivationmentioning

confidence: 99%

“…Human action recognition, and ADL recognition particularly, is an active field of research [6,[8][9][10][11]. For instance, Gomaa et al [8] take data from a smartwatch, compute the autocorrelation function up to a certain lag, then feed these features to a 'random forest'based classifier for training. For their experiments, they use a dataset with 14 different class labels, consisting of a mixture of motion primitives (e.g., walk, lie in bed, stand from sitting), as well as more complex activities (ADLs).…”

Section: Motivationmentioning

confidence: 99%

Privacy-Preserving Human Action Recognition with a Many-Objective Evolutionary Algorithm

Climent-Pérez

Flórez-Revuelta

2022

Sensors

View full text Add to dashboard Cite

Wrist-worn devices equipped with accelerometers constitute a non-intrusive way to achieve active and assisted living (AAL) goals, such as automatic journaling for self-reflection, i.e., lifelogging, as well as to provide other services, such as general health and wellbeing monitoring, personal autonomy assessment, among others. Human action recognition (HAR), and in particular, the recognition of activities of daily living (ADLs), can be used for these types of assessment or journaling. In this paper, a many-objective evolutionary algorithm (MaOEA) is used in order to maximise action recognition from individuals while concealing (minimising recognition of) gender and age. To validate the proposed method, the PAAL accelerometer signal ADL dataset (v2.0) is used, which includes data from 52 participants (26 men and 26 women) and 24 activity class labels. The results show a drop in gender and age recognition to 58% (from 89%, a 31% drop), and to 39% (from 83%, a 44% drop), respectively; while action recognition stays closer to the initial value of 68% (from: 87%, i.e., 19% down).

show abstract

“… [7] 2014 8 19 MPs (sport) Mobifall [8] 2014 24 9 MPs + 4 falls SAR [9] 2014 10 7 MPs mHealth [10] 2015 10 12 MPs (sport) Stisen et al. [11] 2015 9 6 MPs JSI+FoS [12] 2016 15 10 MPs ADLs dataset [13] 2017 – 14 ADLs ASTRI [14] 2019 11 5 MPs Intelligent Fall [15] 2019 6/11 16 ADLs + 5 falls IM-WSHA [16] 2020 10 11 ADLs Fioretti et al. [17] 2021 36 6 ADLs Proposed: PAAL ADL v1.0 [18] 2021 33 24 ADLs Proposed: PAAL ADL v2.0 2021 52 24 ADLs The data can be used simply for recognition of human activities from accelerometer data, or, as proposed by the authors, to also minimise leakage of identity.…”

Section: Value Of the Datamentioning

confidence: 99%