Sensing services for the detection of humans and animals by analyzing the environmental changes of wireless local area network (WLAN) signals have attracted attention in recent years. In object detection using WLAN signals, a widely known technique is the use of time changes in received signal strength indicators that are easily measured between WLAN devices. Utilizing channel response, including power and phase values per subcarrier on multiple input multiple output (MIMO), the orthogonal frequency division multiplexing transmission was researched as channel state information (CSI) to further improve detection accuracy. This paper describes a WLAN-based CSI monitoring system that efficiently acquires the CSI of multiple links in a target area where multiple CSI measuring stations are distributed. In the system, a novel CSI monitoring station captures wireless packets sent within the area and extracts CSI by analyzing the packets on the sounding protocol, specified by IEEE 802.11ac. The paper also describes the system configuration and shows that indoor experimental measurements confirmed the system’s feasibility.
A bathroom has higher probability of accidents than other rooms due to a slippery floor and temperature change. Because of high privacy and humidity, we face difficulties in monitoring inside a bathroom using traditional healthcare methods based on cameras and wearable sensors. In this paper, we present a danger-pose detection system using commodity Wi-Fi devices, which can be applied to bathroom monitoring, preserving privacy. A machine learning-based detection method usually requires data collected in target situations, which is difficult in detection-of-danger situations. We therefore employ a machine learning-based anomaly-detection method that requires a small amount of data in anomaly conditions, minimizing the required training data collected in dangerous conditions. We first derive the amplitude and phase shift from Wi-Fi channel state information (CSI) to extract low-frequency components that are related to human activities. We then separately extract static and dynamic features from the CSI changes in time. Finally, the static and dynamic features are fed into a one-class support vector machine (SVM), which is used as an anomaly-detection method, to classify whether a user is not in bathtub, bathing safely, or in dangerous conditions. We conducted experimental evaluations and demonstrated that our danger-pose detection system achieved a high detection performance in a non-line-of-sight (NLOS) scenario.
The deployment of a large-scale indoor sensor network faces a sensor localization problem because we need to manually locate significantly large numbers of sensors when Global Positioning System (GPS) is unavailable in an indoor environment. Fingerprinting localization is a popular indoor localization method relying on the received signal strength (RSS) of radio signals, which helps to solve the sensor localization problem. However, fingerprinting suffers from low accuracy because of an RSS instability, particularly in sensor localization, owing to low-power ZigBee modules used on sensor nodes. In this paper, we present MuCHLoc, a fingerprinting sensor localization system that improves the localization accuracy by utilizing channel diversity. The key idea of MuCHLoc is the extraction of channel diversity from the RSS of Wi-Fi access points (APs) measured on multiple ZigBee channels through fingerprinting localization. MuCHLoc overcomes the RSS instability by increasing the dimensions of the fingerprints using channel diversity. We conducted experiments collecting the RSS of Wi-Fi APs in a practical environment while switching the ZigBee channels, and evaluated the localization accuracy. The evaluations revealed that MuCHLoc improves the localization accuracy by approximately 15% compared to localization using a single channel. We also showed that MuCHLoc is effective in a dynamic radio environment where the radio propagation channel is unstable from the movement of objects including humans.
The advancement of ICT and affordability of medical sensors enable healthcare data to be obtained remotely. Remote healthcare data is erroneous in nature. Detection of errors for remote healthcare data has not been significantly studied. This research aims to design and develop a software system to detect and reduce such healthcare data errors. Enormous research efforts produced error detection algorithms, however, the detection is done at the server side after a substantial amount of data is archived. Errors can be efficiently reduced if the suspicious data can be detected at the source. We took the approach to predict acceptable range of anthropometric data of each patient. We analyzed 40,391 records to monitor the growth patterns. We plotted the anthropometric items e.g., Height, Weight, BMI, Waist and Hip size for males and females. The plots show some patterns based on different age groups. This paper reports one parameter, height of males. We found three groups that can be classified with similar growth patterns: Age group 20–49, no significant change; Age group 50–64, slightly decremented pattern; and Age group 65–100, a drastic height loss. The acceptable range can change over time. The system estimates the updated trend from new health records.
In this paper, we provide a study on the effects of applying classical clustering algorithms, such as k-Means to free text recommender systems. A typical recommender system may face problems when the number of items from a database goes from a few items to hundreds of items. Currently, one of the most prominent techniques to scale the database is applying clustering, however clustering may have a negative impact on the accuracy of the system when applied without taking into consideration the underlying items. In this work, we build a conceptual text recommender system and use k-Means to partition its search space into different groups. We study how the variation of the number of clusters affects its performance in the light of two performance measurements: recommendation time and precision. We also analyze if this clustering is affected by the representation of text we use. All the techniques used in this study uses word-embeddings to represent the document. One of the main findings of this work is that using clustering we can improve the recommendation time in up to almost 30 times without affecting much off its initial accuracy. Another interesting finding is that the increment of the number of clusters is not directly translated into linear performance.