Evaluation of Smartphone Inertial Sensor Performance for Cross-Platform Mobile Applications

Kos, Anton; Tomažič, Sašo; Umek, Anton

doi:10.3390/s16040477

Cited by 54 publications

(48 citation statements)

References 25 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…And small inertial sensor error can accumulate to a drastically erroneous influence along with time. According to [14], a normal cell phone has about 8.1mrad/s biases in gyroscope and 15mg 0 for X-and Y-axes, 25mg 0 for Z-axes biases in accelerometer measurements. We apply this constant bias error with scale on PoseGT to examine the robustness of our model.…”

Section: Experimental Setupsmentioning

confidence: 99%

Self-supervised Learning of Depth and Camera Motion from 360 $$^\circ $$ Videos

Wang

Cheng

et al. 2019

Lecture Notes in Computer Science

View full text Add to dashboard Cite

As 360 • cameras become prevalent in many autonomous systems (e.g., self-driving cars and drones), efficient 360 • perception becomes more and more important. We propose a novel self-supervised learning approach for predicting the omnidirectional depth and camera motion from a 360 • video. In particular, starting from the SfMLearner, which is designed for cameras with normal field-of-view, we introduce three key features to process 360 • images efficiently. Firstly, we convert each image from equirectangular projection to cubic projection in order to avoid image distortion. In each network layer, we use Cube Padding (CP), which pads intermediate features from adjacent faces, to avoid image boundaries. Secondly, we propose a novel "spherical" photometric consistency constraint on the whole viewing sphere. In this way, no pixel will be projected outside the image boundary which typically happens in images with normal field-of-view. Finally, rather than naively estimating six independent camera motions (i.e., naively applying SfM-Learner to each face on a cube), we propose a novel camera pose consistency loss to ensure the estimated camera motions reaching consensus. To train and evaluate our approach, we collect a new PanoSUNCG dataset containing a large amount of 360 • videos with groundtruth depth and camera motion. Our approach achieves state-of-the-art depth prediction and camera motion estimation on PanoSUNCG with faster inference speed comparing to equirectangular. In real-world indoor videos, our approach can also achieve qualitatively reasonable depth prediction by acquiring model pre-trained on PanoSUNCG.

show abstract

Section: Experimental Setupsmentioning

confidence: 99%

Self-supervised Learning of Depth and Camera Motion from 360 $$^\circ $$ Videos

Wang

Cheng

et al. 2019

Lecture Notes in Computer Science

View full text Add to dashboard Cite

show abstract

“…Deep learning architectures have been used for both feature extraction and activity classification. In recent research, Convolutional Neural Networks (CNNs) [8] and Recurrent Neural Networks (RNNs) [9] have been applied as core of deep learning systems for HAR.…”

Section: Introductionmentioning

confidence: 99%

Human Activity Recognition Based on Deep Learning Techniques

Gil-Martín

Sánchez-Hernández

San-Segundo

2019

The 6th International Electronic Conference on Sensors and Applications

View full text Add to dashboard Cite

Deep learning techniques are being widely applied to Human Activity Recognition (HAR). This paper describes the implementation and evaluation of a HAR system for daily life activities using the accelerometer of an iPhone 6S. This system is based on a deep neural network including convolutional layers for feature extraction from accelerations and fully-connected layers for classification. Different transformations have been applied to the acceleration signals in order to find the appropriate input data to the deep neural network. This study has used acceleration recordings from the MotionSense dataset, where 24 subjects performed 6 activities: walking downstairs, walking upstairs, sitting, standing, walking and jogging. The evaluation has been performed using a subject-wise cross-validation: recordings from the same subject do not appear in training and testing sets at the same time. The proposed system has obtained a 9% improvement in accuracy compared to the baseline system based on Support Vector Machines. The best results have been obtained using raw data as input to a deep neural network composed of two convolutional and two max-pooling layers with decreasing kernel sizes. Results suggest that using the module of the Fourier transform as inputs provides better results when classifying only between dynamic activities.

show abstract

“…Typical examples in the outdoors environment found in the literature include the development of traffic monitoring and prediction systems [7,8], the setup of route choice models [9], the collection of intersection performance data [10], driver behavior studies [11,12], as well as, traffic flow analyses [13,14]. Other studies deal with more specialized transportation topics, such as carbon emission footprint estimation [15] and evaluation of smartphone inertial sensors for cross-platform mobile applications [16]. More recently, crowdsourced smartphone data have also been used in the context of smart cities for emerging mobility applications, including pedestrian mobility management [17], and models to measure the workload associated with the processes of social internet of vehicles [18].…”

Section: Introductionmentioning

confidence: 99%

Rigorous Performance Evaluation of Smartphone GNSS/IMU Sensors for ITS Applications

Gikas

Perakis

2016

Sensors

View full text Add to dashboard Cite

With the rapid growth in smartphone technologies and improvement in their navigation sensors, an increasing amount of location information is now available, opening the road to the provision of new Intelligent Transportation System (ITS) services. Current smartphone devices embody miniaturized Global Navigation Satellite System (GNSS), Inertial Measurement Unit (IMU) and other sensors capable of providing user position, velocity and attitude. However, it is hard to characterize their actual positioning and navigation performance capabilities due to the disparate sensor and software technologies adopted among manufacturers and the high influence of environmental conditions, and therefore, a unified certification process is missing. This paper presents the analysis results obtained from the assessment of two modern smartphones regarding their positioning accuracy (i.e., precision and trueness) capabilities (i.e., potential and limitations) based on a practical but rigorous methodological approach. Our investigation relies on the results of several vehicle tracking (i.e., cruising and maneuvering) tests realized through comparing smartphone obtained trajectories and kinematic parameters to those derived using a high-end GNSS/IMU system and advanced filtering techniques. Performance testing is undertaken for the HTC One S (Android) and iPhone 5s (iOS). Our findings indicate that the deviation of the smartphone locations from ground truth (trueness) deteriorates by a factor of two in obscured environments compared to those derived in open sky conditions. Moreover, it appears that iPhone 5s produces relatively smaller and less dispersed error values compared to those computed for HTC One S. Also, the navigation solution of the HTC One S appears to adapt faster to changes in environmental conditions, suggesting a somewhat different data filtering approach for the iPhone 5s. Testing the accuracy of the accelerometer and gyroscope sensors for a number of maneuvering (speeding, turning, etc.,) events reveals high consistency between smartphones, whereas the small deviations from ground truth verify their high potential even for critical ITS safety applications.

show abstract

Evaluation of Smartphone Inertial Sensor Performance for Cross-Platform Mobile Applications

Cited by 54 publications

References 25 publications

Self-supervised Learning of Depth and Camera Motion from 360 $$^\circ $$ Videos

Self-supervised Learning of Depth and Camera Motion from 360 $$^\circ $$ Videos

Human Activity Recognition Based on Deep Learning Techniques

Rigorous Performance Evaluation of Smartphone GNSS/IMU Sensors for ITS Applications

Contact Info

Product

Resources

About