Among techniques designated for indoor localization, wireless fingerprinting is the most emerging because of the widespread deployment of wireless networks. Moreover, positioning methods based on received signal strength indicator fingerprint are attractive for their accuracy and independence from the radio propagation model. This paper describes an implementation of Bluetooth Low Energy positioning method based on fingerprint technique, according to Wi-Fi localization techniques. Adopting the received signal strength indicator and an accurate model, a localization system with a good accuracy is obtained. The method feature of not limiting the freedom and privacy of users, makes it advisable for elderly behavior monitoring.
Advances in medical diagnostics and treatments have produced lengthening life expectancy and, consequently, rise in elderly people in many nations. In fact, health technology contributes both to disease prevention and to limit the decay of a person functions which corresponds to increase in life expectancy [1-6]. Many elderly live alone in their home because they are creatures of habit. Unfortunately, their deteriorated physical functions severely reduce their mobility and their self-sufficiency. Therefore, elderly are prone to abrupt health problems such as falls, sudden illness, loss of consciousness, etc. Modern advances in the fields of information technology, communication/sensor networks and electronic devices are enabling the development of technological solutions for elderly taking care [7-8]. Many efforts have attempted to monitor the location where old people live by applying intelligent technologies, such as image surveillance, sound detection or vibration sensing for the identification of fall or trip hazards [9, 10]. For this reason, indoor positioning systems have rapidly developed by using new technologies and methods. One of the most popular positioning technologies based on 2D modeling is the fingerprint on Wi-Fi which can be used in both outdoor and indoor environment. Because of Wi-Fi localization system need at least three Wi
Fi routers in the monitoring area, it is not advisable for elderly monitoring in their home. Compared with Wi-Fi localization system, Bluetooth Low Energy (BLE) fingerprinting is a low-cost technology in fact it need only to install beacons in the monitoring area that are battery operated cheap devices. Moreover, these systems are characterized by high precision but the algorithm complexity and the computational consumption is relatively high [11]. Like all indoor positioning systems, fingerprint technology is influenced by the environment such as multipath effects, number of access (AP or BC) and reference points (RP), presence of mobile devices, etc.
BLE fingerprinting is characterized by an off-line phase followed by an on-line phase [12]. During the off-line stage fingerprint data are collected to build a model; in fact the location dependent characteristics of a signal acquired at known locations are stored in a database or radio map. For the model development, both accurate and empirical models can be used [11], [13]. Adopting empirical modeling, the measure of the signal strength received from different beacons (BC) at the location of every RP is performed and stored in the database with the information of RP location. This modeling method has a high level of accuracy even if the workload is, often, large for the need of many data samples. For accurate modeling, the accurate position of BCs is necessary. Moreover, for the radio map generation, the signal strength of several important positions is only needed. Because of complexity and limitations of indoor environment which causes scattering, reflection, and refraction of the propagated RF signal, the main task of accurate modeling is to define a channel propagation model able to describe indoor signal fading properly. To remove invalid samples and improve accuracy of the positioning method, the fingerprint database has to be filtered since the acquired signal strength is corrupted by noise generated by the indoor environment. Both deterministic denoising and probabilistic denoising methods can be used.
In the online phase, collected signal strengths are compared to the signal strength at the user location to estimate the user position adopting deterministic or probabilistic positioning algorithms. These type of procedures differ mostly because deterministic algorithms match the signal strength of corresponding fingerprint data by using the positioning algorithm to obtain the final position of the user, while probabilistic algorithms store the probability distribution of the signal strength during a certain time in the fingerprint database and then the probability position of the user location is evaluated by the Bayesian theory system.
In this paper, a monitoring system for elderly people based on BLE technology is presented. The method performs indoor positioning adopting Wi-Fi fingerprinting technique based on Received Signal Strength Indicator (RSSI). After a brief description of the BLE technology, the implemented system for the indoor localization is detailed. The method performance is evaluated and some conclusions are drawn out. 2
Bluetooth low energy technology combines a standardized technology designed for ultra-low-power batteries and a new sensor-based data collection framework. It boomed in the mobile device market and uses devices working at 2.4 GHz and within the Industrial, Scientific and Medical (ISM) band. BLE technology employs a variable connection interval that can be set from a few milliseconds to several seconds depending on the application. In addition, because it features a very rapid connection, BLE technology can normally be in a "not connected" state (saving power) where the two ends of a link are aware of each other, but only link up when absolutely necessary and then for as short a time as possible. Benefits of BLE technology are [14]: • Low power consumption: nowadays, in order to reduce the cost of system infrastructure, the most important requirement for all electronic systems is a low power consumption; • Indoor positioning: BLE technology is designed specifically for this purpose; • Supported by smartphones: this feature is very important in view of trends in the mobile telephone systems; • Very long life battery: this is a very important aspect for all portable systems. A long life battery enables a long life system operation and, consequently, the possibility to replace batteries after a few months of use. As a result, system maintenance costs decrease; • Low cost: cost control is one of the requirement/constraint in designing an electronic system. The adoption of low-cost hardware guarantees cost reduction for system implementation. 3
The purpose of the implemented system is to localize the positions of one receiver (i.e. smartphone) in the physical area of interest using the BLE technology. In BLE fingerprint technique, the definition of a radio frequency map is necessary which combines geographical coordinates (2-Dimension Cartesian Space) and RSSI values received by the tag and transmitted by several beacons (BC). The location of beacons is a key point because it must be chosen in such a way that at any time the tag is within the radio range of at least one beacon. The generic RSSI vector received by a receiver is denoted as t = (t1, t2,…, tn), where tj, (1≤j≤n) denotes the RSSI value from the jth beacon and n is the number of beacons in the area of interest. Adopting m RPs in the offline phase, the RSSI vector received at the ith RP (1≤i≤m) is denoted as si, where si = (si1, si2,…sin).
The implemented method is composed of an off-line phase, in which the fingerprint data base is constructed adopting an accurate model, and an on-line phase in which a deterministic positioning algorithm is implemented to localize the receiver inside the area under test.
In the implemented system, four beacons and a mobile receiver are used inside an indoor environment of about 133 m2 (fig.1).
In order to reduce the effort dedicated to off-line measurements, a virtual fingerprint database is calculated using the Multi-Wall Multi-Floor indoor propagation model whose effectiveness was investigated [15]. The RSSI database is composed of 250000 elements (m=250000) (fig 1). To simulate the RSSI signals measured by the receiver if it is at the jth RP, the sj vector was corrupted by random values (noise) which are different for all its components. As previously mentioned, the signals read by the receiver is called t, where t = (t1, t2,…, tn) and ti, = sij +cij.
This assumption is justified because errors can corrupt the RSSI values measured by the receiver. The Mersenne Twister algorithm is adopted for the generation of random values. In particular,
cij = (b-a) × rand (1) + a (1) where a and b are the boundary of the signal corrupted by the error a = sij-MaxErr × MSV (2) b = sij+ MaxErr × MSV (3) and Maximum Scale Value(MSV) is the maximum signal value assumed to be -25dB. MaxErr is in the range [0,1], assuming that 100% is the maximum read error that can be done by the receiver.
For software implementation (online phase), all the sj points are evaluated with the aim to determine if the sj point is a probably matching point or not.
The criteria adopted is based on the successful condition that each value of t is in sj range that is for each sj, ti has to be in sij range. The range sij is a window of size equal to 50% of the ti value and centered around si value.
For j=1..m , sj is a matching point for t if:
1 − 0.5 × 1 ≤ 1 ≤ 1 + 0.5 × 1 { The implemented procedure localizes all the s points inside the area under test which should represent the receiver position.
In fig.2, the receiver real position is indicated by a circle while the gray areas represent the receiver localizations obtained by the implemented procedure. In the figure is easily identify k cluster of probable points. The second step consists to evaluating the numerical consistence of each cluster. Only the cluster with the maximum point consistence is considered in the next step. The third step is the evaluation of the cluster centroid.
For an accurate positioning of the receiver, the k-means clustering algorithm with the Euclidean distance measurement technique, is adopted. The used procedure partitions a set of n objects into k clusters so that the resulting intra-cluster similarity is high but the inter-cluster similarity is low. The procedure defines one centroid for each cluster. Generally speaking, a centroid is an artificial point in the space of records which represents an average location of the particular cluster. Its coordinates are averages of attribute values of all examples that belong to the cluster.
The k-means algorithm uses an heuristic method to find centroid seeds for k-means clustering. Assuming k different clusters, the algorithm chooses the seeds as follows:
It selects an observation uniformly at random from the data set, X. The chosen observation is the first centroid, and is denoted with 1. 2. It computes the distances from each observation to 1 and denotes the distance between and the observation m with d( , ). 3. It selects the next centroid 2 at random from X with probability 2( , 1) ∑ =1 2( , 1)
a) b)
Computes the distances from each observation to each centroid and assigns each observation to its closest centroid.
Selects centroid j at random from X with probability
2( , ) ∑{ℎ, ℎ∈ }
2( ℎ, ) for m = 1, …, n and p = 1, …, j-1 where is the set of all observations closest to centroid and belongs to . That is, the algorithm selects each subsequent center with a probability proportional to the distance from itself to the closest center already chosen. 5.
Repeat step 4 until k centroids are chosen.
The estimated localization point is considered the cluster centroid.
For the evaluation of algorithm robustness towards noise (such as effects of building layout, construction material, moving objects, reflecting surfaces, temperature changes day, electromagnetic field interference, etc.), several simulations were carried. Assuming a transmission power level of each beacon equal to -25dBm, average error distance values less than 1,4m are obtained for noise values not exceeding 20% of maximum scale value (MaxErr=0.2) (fig.3). Circle = real position
Square= simulated position Fig.3 Results obtained with a noise level of 5% (a), 20% (b) and 40% (c) of the maximum scale value (MaxErr=0.05, 0.2, 0.4). 4
In this paper, a BLE fingerprint method using Received Signal Strength Indicator for indoor localization is presented. In the off-line phase an accurate model is adopted for the radio map construction, in the on-line phase a deterministic positioning algorithm is implemented to localize the receiver inside the test area. The obtained results show the method validity.
References 1. Rizzi M., D’Aloia M., Guaragnella C., Castagnolo B., Health Care Improvement: Comparative Analysis of Two CAD Systems in Mammographic Screening, IEEE Trans. Syst., Man, Cybern. A, Syst., Humans 42, 1385-1395 (2012) 2. Yuan Y., Giger M.L., Li H., Bhooshan N., Sennett C.A., Correlative Analysis of FFDM and
DCE-MRI for Improved Breast CADx, J. Med. Biol. Eng., 32, 42-50 (2012). 3. Rizzi M., D’Aloia M., Computer Aided System for Breast Cancer Diagnosis, Biomed. Eng
App. Bas C. 26,1450033 (2014) 4. Rizzi M., D’Aloia M., Cice G., Computer aided evaluation (CAE) of morphologic changes in pigmented skin lesions, Lect. Notes Comput. Sc. (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) 9281, 250-257 (2015) 5. Pirlo G., Impedovo D., On the measurement of local stability of handwriting: An application to static signature verification, 2010 IEEE Workshop on Biometric Measurements and Systems for Security and Medical Applications, 41-44, (2010) 6. Rizzi M., D’Aloia M., Castagnolo B., Semiconductor detectors and principles of radiationmatter interaction. Journal of Applied Sciences 10, 3141-3155 (2010) 7. Hsu C.C., Chen J.H., A Novel Sensor-Assisted RFID-Based Indoor Tracking System for the
Elderly Living Alone, Sensors 11, 10094-10113 (2012) 8. Impedovo D., Pirlo G., Updating Knowledge in Feedback-Based Multi-classifier Systems, 2011 Int. Conference on Document Analysis and Recognition, Beijing, 227-231 (2011) 9. Bakar U.A.B.U.A., Ghayvat H., Hasanm S.F., Mukhopadhyay S.C., Activity and Anomaly Detection in Smart Home: A Survey. In: Mukhopadhyay S. (eds) Next Generation Sensors and Systems. Smart Sensors, Measurement and Instrumentation 16. Springer, Cham (2016) 10. D'Aloia M., Cortone F., Cice G., Russo R., Rizzi M., Longo A., Improving energy efficiency in building system using a novel people localization system, 2016 IEEE Workshop on Environmental, Energy, and Structural Monitoring Systems, EESMS 2016; Bari (Italy), 13-14 June (2016) 11. Xia S., Liu Y., Yuan G., Zhu M., Wang Z., Indoor fingerprint positioning based on Wi-Fi: an overview, ISPRS Int. J. Geo-Inf. 6, 135, (2017) 12. Suining H., Gary Chan S. H., Wi-Fi Fingerprint-Based Indoor Positioning: Recent Advances and Comparisons, IEEE Commun. Surv Tut, 18, 466-490 (2016) 13. Swangmuang N.; Krishnamurthy P.V., On clustering RSS fingerprints for improving scalability of performance prediction of indoor positioning systems, 1st ACM Int. Work. on Mobile Entity Localization and Tracking in GPS-Less Environments, San Francisco, (USA), 61-66 (2008) 14. Lee J.S., Dong M. F., Sun Y.H., A preliminary study of low power wireless technologies: ZigBee and Bluetooth Low Energy, 10th IEEE Conf. on. Industrial Electronics and Applications (ICIEA), (2015) 15. Caso G., De Nardis L., On the Applicability of Multi-wall Multi-floor Propagation Models to WiFi Fingerprinting Indoor Positioning, Future Access Enablers for Ubiquitous and Intelligent Infrastructures: Lecture Notes of the Institute for Computer Sciences, Social Informatics and Telecommunications Engineering, Springer, (2015)