Convolutional Neural Networks for Human Activity Recognition using Mobile Sensors

Abstract

A variety of real-life mobile sensing applications are becoming available, especially in the life-logging, fitness tracking and health monitoring domains. These applications use mobile sensors embedded in smart phones to recognize human activities in order to get a better understanding of human behavior. While progress has been made, human activity recognition remains a challenging task. This is partly due to the broad range of human activities as well as the rich variation in how a given activity can be performed. Using features that clearly separate between activities is crucial. In this paper, we propose an approach to automatically extract discriminative features for activity recognition. Specifically, we develop a method based on Convolutional Neural Networks (CNN), which can capture local dependency and scale invariance of a signal as it has been shown in speech recognition and image recognition domains. In addition, a modified weight sharing technique, called partial weight sharing, is proposed and applied to accelerometer signals to get further improvements. The experimental results on three public datasets, Skoda (assembly line activities), Opportunity (activities in kitchen), Actitracker (jogging, walking, etc.), indicate that our novel CNN-based approach is practical and achieves higher accuracy than existing state-of-the-art methods.

Wearable Inertial Sensor Systems for Lower Limb Exercise Detection and Evaluation: A Systematic Review

“…Future work may also benefit from utilising deep learning techniques for classification such as the convolutional neural networks approach demonstrated by Veiga et al [68]. Such classification methodologies have recently been shown to have many benefits when compared to traditional machine learning classification techniques when analysing timeseries data, including reducing the risk of overfitting and improving system accuracy [77,78] (Table 10) [3,19,23,64,[70][71][72]74]. There is therefore potential to investigate movement classification with larger data sets and across a range of other exercises.…”

Can Deep Learning Revolutionize Mobile Sensing?

“…Some existing application domains of deep learning (such as emotion recognition [15] and others related to audio) are very similar to requirements of mobile sensing and should be able to be adapted for sensor app purposes. Other important sensing tasks like activity recognition are largely unexplored in terms of deep learning, with only isolated examples being available (such as for feature selection [24] or non-mobile activity recognition in controlled or instrumented environments [12,3]). These inference tasks will require more fundamental study as they lack clear analogs in the deep learning literature.…”

Deep learning for human activity recognition: A resource efficient implementation on low-power devices

“…However, current methods typically achieve this by introducing additional layers and nodes for classification, which increases computational complexity. For example, in the CNN based method described by Zeng et al [10], additional max-pooling layers are applied after feature detection from the raw input to produce scaleinvariant features, which is then introduced to a 1024 neuron hidden layer to merge features from multiple channels, and another additional soft-max layer is used to generate the classification result. Yang et al [11] and Chen and Xue [12] both use CNNs with multiple iterations of convolution and subsampling layers, or convolution and pooling layers being applied for feature extraction.…”

“…In deep learning, features are abstracted automatically from the data instead of being handcrafted, which allows these machine learning methods to be more effectively used across a range of different classification tasks [10]- [13]. This automatic feature extraction paradigm is also becoming increasingly relevant in the area of Body Sensor Networks (BSN) [14] as sensors are able to generate evergrowing amounts of data, which makes developing handcrafted features a challenging task.…”