Improving Resource Efficiency of Deep Activity Recognition via Redundancy Reduction

Leite, Clayton Frederick Souza; Xiao, Yu

doi:10.1145/3376897.3377859

Cited by 5 publications

(6 citation statements)

References 6 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…However, 6 of them are rarely present in the data. Following previous works [10,24,25], to avoid a heavily imbalanced training, only the remaining 12 activities are used in this work: lying quietly, sitting, standing, ironing, vacuum cleaning, ascending stairs, descending stairs, walking, Nordic walking, bicycling, running, and rope jumping. Transient activities (denoted as the null class) are discarded.…”

Section: Experimental Setup 41 Datasetsmentioning

confidence: 99%

“…The data were collected at 33 Hz from 4 participants equipped with sensors. In our experiments, only the sensory readings from the upper limbs, back, and both feet were considered (as in [10,24,25]). These readings come from 29 different sensor channels (Fig.…”

Section: Experimental Setup 41 Datasetsmentioning

confidence: 99%

“…Each window was labeled with the activity which lasts for the longest within the window. For the PAMAP2, following [11,19,24,25,33], the window size contained approximately 5.12 seconds with 78% overlap. These numbers are 1 and 50% for Skoda, Daphnet, MHEALTH, and Opportunity (also following [11,24,25]).…”

Section: Data Pre-processingmentioning

confidence: 99%

“…As for PAMAP2, we utilize the data from subject 5 for the validation set, subject 6 for the testing set, and the remaining subjects for the training set. This protocol is followed in diverse works [10,11,25,28]. Also following [10,11,25,28], for the Opportunity dataset, the validation set is composed of run 2 from subject 1, and the test set consists of runs 4 and 5 from subjects 2 and 3.…”

Section: Data Pre-processingmentioning

confidence: 99%

“…Among the methods for HAR, deep learning (DL) has been delivering impressive performance without the need for domain knowledge for feature engineering. However, the utilization of DL in real-time requires large amounts of available memory and computational power, considerably beyond the specifications of common low-cost off-the-shelf microcontrollers [25]. The computational overhead typically increases with the number of sensor channels involved in the input.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Optimal sensor channel selection for resource-efficient deep activity recognition

Leite

Xiao

2021

Proceedings of the 20th International Conference on Information Processing in Sensor Networks (Co-Located With CPS-IoT Week 202

Self Cite

View full text Add to dashboard Cite

Deep learning has permitted unprecedented performance in sensorbased human activity recognition (HAR). However, deep learning models often present high computational overheads, which poses challenges to their implementation on resource-constraint devices such as microcontrollers. Usually, the computational overhead increases with the input size. One way to reduce the input size is by constraining the number of sensor channels. We refer to sensor channel as a specific data modality (e.g. accelerometer) placed on a specific body location (e.g. chest). Identifying and removing irrelevant and redundant sensor channels is feasible via exhaustive search only in cases where few candidates exist. In this paper, we propose a smarter and more efficient way to optimize the sensor channel selection during the training of deep neural networks for HAR. Firstly, we propose a light-weight deep neural network architecture that learns to minimize the use of redundant and irrelevant information in the classification task, while achieving high performance. Secondly, we propose a sensor channel selection algorithm that utilizes the knowledge learned by the neural network to rank the sensor channels by their contribution to the classification task. The neural network is then trimmed by removing the sensor channels with the least contribution from the input and pruning the corresponding weights involved in processing them. The pipeline that consists of the above two steps iterates until the optimal set of sensor channels has been found to balance the trade-off between resource consumption and classification performance. Compared with other selection methods in the literature, experiments on 5 public datasets showed that our proposal achieved significantly higher F1-scores at the same time as utilizing from 76% to 93% less memory, with up to 75% faster inference time and as far as 76% lower energy consumption. CCS CONCEPTS• Human-centered computing → Ubiquitous and mobile computing; • Computing methodologies → Neural networks.

show abstract

Section: Experimental Setup 41 Datasetsmentioning

confidence: 99%

Section: Experimental Setup 41 Datasetsmentioning

confidence: 99%