A Fast Indoor/Outdoor Transition Detection Algorithm Based on Machine Learning

Current navigation systems use multi-sensor data to improve the localization accuracy, but often without certitude on the quality of those measurements in certain situations. The context detection will enable us to build an adaptive navigation system to improve the precision and the robustness of its localization solution by anticipating possible degradation in sensor signal quality (GNSS in urban canyons for instance or camera-based navigation in a non-textured environment). That is why context detection is considered the future of navigation systems. Thus, it is important firstly to define this concept of context for navigation and to find a way to extract it from available information. This paper overviews existing GNSS and on-board vision-based solutions of environmental context detection. This review shows that most of the state-of-the art research works focus on only one type of data. It confirms that the main perspective of this problem is to combine different indicators from multiple sensors.

Section: Combination Of Multiple Indicatorsmentioning

confidence: 99%

Section: Combination Of Multiple Indicatorsmentioning

confidence: 99%

See 1 more Smart Citation

A Review of Environmental Context Detection for Navigation Based on Multiple Sensors

Feriol

Vivet

Watanabe

2020

“…Scanning a QR code may be more user-friendly approach compared to a manual position setup. An outdoor–indoor transition may be detected using machine learning approaches [ 26 ]. When the navigation route or the localization initialization starts at the entrance of the building, the initialization with GNSS signal is possible.…”

Section: Solution Background and Related Workmentioning

confidence: 99%

Grid-Based Bayesian Filtering Methods for Pedestrian Dead Reckoning Indoor Positioning Using Smartphones

Opiela

Galčík

2020

Indoor positioning systems for smartphones are often based on Pedestrian Dead Reckoning, which computes the current position from the previously estimated location. Noisy sensor measurements, inaccurate step length estimations, faulty direction detections, and a demand on the real-time calculation introduce the error which is suppressed using a map model and a Bayesian filtering. The main focus of this paper is on grid-based implementations of Bayes filters as an alternative to commonly used Kalman and particle filters. Our previous work regarding grid-based filters is elaborated and enriched with convolution mask calculations. More advanced implementations, the centroid grid filter, and the advanced point-mass filter are introduced. These implementations are analyzed and compared using different configurations on the same raw sensor recordings. The evaluation is performed on three sets of experiments: a custom simple path in faculty building in Slovakia, and on datasets from IPIN competitions from a shopping mall in France, 2018 and a research institute in Italy, 2019. Evaluation results suggests that proposed methods are qualified alternatives to the particle filter. Advantages, drawbacks and proper configurations of these filters are discussed in this paper.

“…State-of-the-art solutions for pedestrian activity recognition can be divided into two categories: traditional methods and deep learning-based approaches [ 12 , 13 ]. Traditional methods usually consist of two parts: feature extraction and classification.…”

Section: Introductionmentioning

confidence: 99%

Deep Learning-Based Human Activity Real-Time Recognition for Pedestrian Navigation

Zhang

et al. 2020

Several pedestrian navigation solutions have been proposed to date, and most of them are based on smartphones. Real-time recognition of pedestrian mode and smartphone posture is a key issue in navigation. Traditional ML (Machine Learning) classification methods have drawbacks, such as insufficient recognition accuracy and poor timing. This paper presents a real-time recognition scheme for comprehensive human activities, and this scheme combines deep learning algorithms and MEMS (Micro-Electro-Mechanical System) sensors’ measurements. In this study, we performed four main experiments, namely pedestrian motion mode recognition, smartphone posture recognition, real-time comprehensive pedestrian activity recognition, and pedestrian navigation. In the procedure of recognition, we designed and trained deep learning models using LSTM (Long Short-Term Memory) and CNN (Convolutional Neural Network) networks based on Tensorflow framework. The accuracy of traditional ML classification methods was also used for comparison. Test results show that the accuracy of motion mode recognition was improved from 89.9 % , which was the highest accuracy and obtained by SVM (Support Vector Machine), to 90.74 % (LSTM) and 91.92 % (CNN); the accuracy of smartphone posture recognition was improved from 81.60 % , which is the highest accuracy and obtained by NN (Neural Network), to 93.69 % (LSTM) and 95.55 % (CNN). We give a model transformation procedure based on the trained CNN network model, and then obtain the converted . t f l i t e model, which can be run in Android devices for real-time recognition. Real-time recognition experiments were performed in multiple scenes, a recognition model trained by the CNN network was deployed in a Huawei Mate20 smartphone, and the five most used pedestrian activities were designed and verified. The overall accuracy was up to 89.39 % . Overall, the improvement of recognition capability based on deep learning algorithms was significant. Therefore, the solution was helpful to recognize comprehensive pedestrian activities during navigation. On the basis of the trained model, a navigation test was performed; mean bias was reduced by more than 1.1 m. Accordingly, the positioning accuracy was improved obviously, which is meaningful to apply DL in the area of pedestrian navigation to make improvements.