Location Adaptive Motion Recognition Based on Wi-Fi Feature Enhancement
0
Citation
37
Reference
10
Related Paper
Abstract:
Action recognition is essential in security monitoring, home care, and behavior analysis. Traditional solutions usually leverage particular devices, such as smart watches, infrared/visible cameras, etc. These methods may narrow the application areas due to the risk of privacy leakage, high equipment cost, and over/under-exposure. Using wireless signals for motion recognition can effectively avoid the above problems. However, the motion recognition technology based on Wi-Fi signals currently has some defects, such as low resolution caused by narrow signal bandwidth, poor environmental adaptability caused by the multi-path effect, etc., which make it hard for commercial applications. To solve the above problems, we first propose and implement a position adaptive motion recognition method based on Wi-Fi feature enhancement, which is composed of an enhanced Wi-Fi features module and an enhanced convolution Transformer network. Meanwhile, we improve the generalization ability in the signal processing stage to avoid building an extremely complex model and reduce the demand for system hardware. To verify the generalization of the method, we implement real-world experiments using 9300 network cards and the PicoScenes software platform for data acquisition and processing. By contrast with the baseline method using original channel state information(CSI) data, the average accuracy of our algorithm is improved by 14% in different positions and over 16% in different orientations. Meanwhile, our method has best performance with an accuracy of 90.33% compared with the existing models on public datasets WiAR and WiDAR.Keywords:
Leverage (statistics)
Activity Recognition
오늘날 스마트폰의 발전으로 스마트폰 내장 센서를 통해 사용자의 개인 정보를 쉽게 파악 할 수 있고 원한다면 사용자의 위치를 실시간으로 알아낼 수 있다. 그리하여 센서를 통해 추출된 데이터를 통해 동작인식과 생활 패턴 인식에 관한 연구가 급증하고 있다. 본 논문에서는 기존의 동작 인식 연구에서 추출되는 데이터를 정형화하기 위해 동작 데이터를 모델링하였다. 본 논문의 일상 동작 모델링은 이론적 분석이다. 동작을 크게 두 가지로 분류시켜 가속도 센서만으로 인식 가능한 기본 동작을 물리적 동작으로 정의하고 그 외 목적과 대상, 장소를 포함하는 모든 동작을 논리적 동작으로 분류시켰다. 모델링 된 데이터를 기반으로 각 동작의 특성에 맞게 가시화 하는 방안을 제안하였다. 본 연구를 통해 인간의 일상생활을 동작별로 간편하게 표준화 할 수 있고 기존의 동작 인식 연구에서 추출되는 동작 데이터를 사용자의 요구에 따라 가시화 할 수 있다. With the development of Smartphone, Smartphone contains diverse functions including many sensors that can describe users' state. So there has been increased studies rapidly about activity recognition and life pattern recognition with Smartphone sensors. This research suggest modeling of the activity data to classify extracted data in existing activity recognition study. Activity data is divided into two parts: Physical activity and Logical Activity. In this paper, activity data modeling is theoretical analysis. We classified the basic activity(walking, standing, sitting, lying) as physical activity and the other activities including object, target and place as logical activity. After that we suggested a method of visualizing modeling data for users. Our approach will contribute to generalize human's life by modeling activity data. Also it can contribute to visualize user's activity data for existing activity recognition study.
Activity Recognition
Logical data model
Cite
Citations (5)
Understanding Activities of Human Daily Life is a fundamental and essential AI problem for Ubiquitous Computing and Human-Computer Interaction. Activity inference has attracted enormous research on activity recognition from mobile sensor data. However, it is not clear how different signals can influence activity inference. To this end, we investigated the problem of activity recognition and prediction. Experiments showed that contextual signals like time, location, previous activity and related person are much more useful than demographical signals for activity recognition and prediction. We improved the accuracy of activity recognition by more than 15% comparing to existing work on the same dataset. What's more, we revealed that we can predict what will you do next with high accuracy.
Activity Recognition
Cite
Citations (0)
Activity recognition in smart environments is an important technology for assisted living and e-health. Recently there are growing interests in applying machine learning algorithms to activity recognition tasks. In this paper, we combine support vector machine (SVM) and association rule learning to improve the performance of activity recognition based on streaming sensor data in smart homes. The proposed approach allows us to accurately identify the activity boundaries, hence reducing activity recognition errors in the system.
Activity Recognition
Smart environment
Cite
Citations (5)
Human activity recognition based on smart phones has been widely used in many fields including the mobile context awareness and inertial positioning. Compared to the activity recognition whose sensor location is fixed, the activity recognition based on smartphones has a new problem because the mobile direction and position are not fixed. In this paper, we study the activity recognition on the Android smartphones to find out a location-free method. Firstly, this paper analyzes the human motion and experimental data, and proposes a method using the similarity of activity to achieve the location independence to further improve the recognition precision. Secondly we describe how to calculate and use the similarity in the process of activity recognition to help our research. Finally, the experiments are introduced, including the collection of experimental data, results of different methods and the direction of further study.
Activity Recognition
Similarity (geometry)
Activity detection
Cite
Citations (9)
Activity recognition has become of great importance in many fields especially in fitness monitoring; health and elder care by offering the opportunity for large amount of applications which recognize human's daily life activities. The prevalence of smart phones in our society with their ever growing sensing power has opened the door for more sophisticated data mining applications which takes the raw sensor data as input and classify the motion activity performed. The main sensor used in performing activity recognition is the accelerometer. This paper presents a framework for activity recognition using smart phone sensors. Features extracted from raw sensor data are used to train and test supervised machine learning algorithms.
Activity Recognition
Motion sensors
Activity detection
Mobile phone
Intelligent sensor
Smartwatch
Cite
Citations (0)
Activity Recognition
Activity detection
Activity monitor
Cite
Citations (62)
Activity Recognition
Sensor Fusion
Cite
Citations (13)
This paper describes the results of experiments where information about places is used in the recognition of activities in the home. We explore the use of place-specific activity recognition trained with supervised learning, coupled with a decision fusion step, for recognition of activities in the Opportunity dataset. Our experiments show that using place information to control recognition can substantially improve both the error rates and the computation cost of activity recognition compared to classical approaches where all sensors are used and all activities are possible. The use of place information for controlling recognition gives an F1 classification score of 92:70% ± 1:26%, requiring on average only 73 milliseconds of computing time per instance of activity. These experiments demonstrate that organizing activity recognition with place-based context models can provide a scalable approach for building context-aware services in smart home environments.
Activity Recognition
Sensor Fusion
Home Automation
Cite
Citations (0)
Leverage (statistics)
Media Coverage
Argument (complex analysis)
Cite
Citations (69)