Combining skeleton and accelerometer data for human fine-grained activity recognition and abnormal behaviour detection with deep temporal convolutional networks

2021 ◽  
Author(s):  
Cuong Pham ◽  
Linh Nguyen ◽  
Anh Nguyen ◽  
Ngon Nguyen ◽  
Van-Toi Nguyen
Keyword(s):  
Activity Recognition ◽  
Accelerometer Data ◽  
Abnormal Behaviour ◽  
Fine Grained ◽  
Convolutional Networks

The Effect of Sensor Placement and Number on Physical Activity Recognition and Energy Expenditure Estimation (Preprint)

2020 ◽  
Author(s):  
Anis Davoudi ◽  
Mamoun T. Mardini ◽  
Dave Nelson ◽  
Fahd Albinali ◽  
Sanjay Ranka ◽  
...  
Keyword(s):  
Physical Activity ◽  
Older Adults ◽  
Energy Expenditure ◽  
Activity Recognition ◽  
Recognition Accuracy ◽  
Body Position ◽  
Computer Work ◽  
Accelerometer Data ◽  
Upper Arm ◽  
Lifestyle Activities

BACKGROUND Research shows the feasibility of human activity recognition using Wearable accelerometer devices. Different studies have used varying number and placement for data collection using the sensors. OBJECTIVE To compare accuracy performance between multiple and variable placement of accelerometer devices in categorizing the type of physical activity and corresponding energy expenditure in older adults. METHODS Participants (n=93, 72.2±7.1 yrs) completed a total of 32 activities of daily life in a laboratory setting. Activities were classified as sedentary vs. non-sedentary, locomotion vs. non-locomotion, and lifestyle vs. non-lifestyle activities (e.g. leisure walk vs. computer work). A portable metabolic unit was worn during each activity to measure metabolic equivalents (METs). Accelerometers were placed on five different body positions: wrist, hip, ankle, upper arm, and thigh. Accelerometer data from each body position and combinations of positions were used in developing Random Forest models to assess activity category recognition accuracy and MET estimation. RESULTS Model performance for both MET estimation and activity category recognition strengthened with additional accelerometer devices. However, a single accelerometer on the ankle, upper arm, hip, thigh, or wrist had only a 0.03 to 0.09 MET increase in prediction error as compared to wearing all five devices. Balanced accuracy showed similar trends with slight decreases in balanced accuracy for detection of locomotion (0-0.01 METs), sedentary (0.13-0.05 METs) and lifestyle activities (0.08-0.04 METs) compared to all five placements. The accuracy of recognizing activity categories increased with additional placements (0.15-0.29). Notably, the hip was the best single body position for MET estimation and activity category recognition. CONCLUSIONS Additional accelerometer devices only slightly enhance activity recognition accuracy and MET estimation in older adults. However, given the extra burden of wearing additional devices, single accelerometers with appropriate placement appear to be sufficient for estimating energy expenditure and activity category recognition in older adults.

scholarly journals Detection of foraging behavior from accelerometer data using U-Net type convolutional networks

2021 ◽  
pp. 101275
Author(s):  
Mạnh Cường Ngô ◽  
Raghavendra Selvan ◽  
Outi Tervo ◽  
Mads Peter Heide-Jørgensen ◽  
Susanne Ditlevsen
Keyword(s):  
Foraging Behavior ◽  
Accelerometer Data ◽  
Convolutional Networks

scholarly journals Recurrent Neural Network for Human Activity Recognition in Embedded Systems Using PPG and Accelerometer Data

Electronics ◽  
2021 ◽  
Vol 10 (14) ◽  
pp. 1715
Author(s):  
Michele Alessandrini ◽  
Giorgio Biagetti ◽  
Paolo Crippa ◽  
Laura Falaschetti ◽  
Claudio Turchetti
Keyword(s):  
Neural Network ◽  
Embedded System ◽  
Activity Recognition ◽  
Human Activity ◽  
Recurrent Neural Network ◽  
Low Cost ◽  
Low Complexity ◽  
Human Activity Recognition ◽  
Motion Artifacts ◽  
Accelerometer Data

Photoplethysmography (PPG) is a common and practical technique to detect human activity and other physiological parameters and is commonly implemented in wearable devices. However, the PPG signal is often severely corrupted by motion artifacts. The aim of this paper is to address the human activity recognition (HAR) task directly on the device, implementing a recurrent neural network (RNN) in a low cost, low power microcontroller, ensuring the required performance in terms of accuracy and low complexity. To reach this goal, (i) we first develop an RNN, which integrates PPG and tri-axial accelerometer data, where these data can be used to compensate motion artifacts in PPG in order to accurately detect human activity; (ii) then, we port the RNN to an embedded device, Cloud-JAM L4, based on an STM32 microcontroller, optimizing it to maintain an accuracy of over 95% while requiring modest computational power and memory resources. The experimental results show that such a system can be effectively implemented on a constrained-resource system, allowing the design of a fully autonomous wearable embedded system for human activity recognition and logging.

Multimodal Wearable Sensing for Fine-Grained Activity Recognition in Healthcare

2015 ◽  
Vol 19 (5) ◽  
pp. 26-35 ◽  
Author(s):  
Debraj De ◽  
Pratool Bharti ◽  
Sajal K. Das ◽  
Sriram Chellappan
Keyword(s):  
Activity Recognition ◽  
Fine Grained ◽  
Wearable Sensing

Novel approaches to human activity recognition based on accelerometer data

2018 ◽  
Vol 12 (7) ◽  
pp. 1387-1394 ◽  
Author(s):  
Artur Jordao ◽  
Leonardo Antônio Borges Torres ◽  
William Robson Schwartz
Keyword(s):  
Activity Recognition ◽  
Human Activity ◽  
Human Activity Recognition ◽  
Accelerometer Data ◽  
Novel Approaches

Motion and Location-Based Online Human Daily Activity Recognition

2018 ◽  
pp. 277-296
Author(s):  
Chun Zhu ◽  
Weihua Sheng
Keyword(s):  
Activity Recognition ◽  
Daily Activity ◽  
Human Subject ◽  
Viterbi Algorithm ◽  
Location Information ◽  
Motion Data ◽  
Fine Grained ◽  
Step Algorithm ◽  
Two Phases ◽  
Moderate Cost

In this chapter, the authors propose an approach to indoor human daily activity recognition that combines motion data and location information. One inertial sensor is worn on the thigh of a human subject to provide motion data while a motion capture system is used to record the human location information. Such a combination has the advantage of significantly reducing the obtrusiveness to the human subject at a moderate cost of vision processing, while maintaining a high accuracy of recognition. The approach has two phases. First, a two-step algorithm is proposed to recognize the activity based on motion data only. In the coarse-grained classification, two neural networks are used to classify the basic activities. In the fine-grained classification, the sequence of activities is modeled by a Hidden Markov Model (HMM) to consider the sequential constraints. The modified short-time Viterbi algorithm is used for real-time daily activity recognition. Second, to fuse the motion data with the location information, Bayes' theorem is used to refine the activities recognized from the motion data. The authors conduct experiments in a mock apartment, and the obtained results prove the effectiveness and accuracy of the algorithms.

Sign in / Sign up

Export Citation Format

Share Document