intelligent services to the resident.
Within this framework, the smart home server has the following functions. 1) The virtual map generator makes a virtual map of the smart home (generating a virtual map), and writes the location information of the resident, which is received from a room terminal, on the virtual map (writing the resident?s location). Then, it makes a moving trajectory of the resident by connecting the successive locations of the resident (tracking the resident?s movement). 2) The home appliance controller transmits control commands to home appliances via the home network system to provide intelligent services to the resident. 3) The moving pattern predictor saves the current movement trajectory of the resident, the current action of home appliances, and parameters reflecting the current home environment such as the time, temperature, humidity, and illumination. After storing sufficient information, it may be possible to offer human-oriented intelligent services in which the home appliances spontaneously provide services to satisfy human needs. For example, if the smart home server “knows” that the resident normally wakes up at 7:00 A.M. and takes a shower, it may be possible to turn on the lamps and some music. In addition, the temperature of the shower water can be set automatically for the resident.
2.2 Location-recognition algorithm
In order to determine the location of a resident within a room, an array of PIR sensors are used as shown in Fig. 3. In the figure, the sensing area of each PIR sensor is shown as a circle, and the sensing areas of two or more sensors overlap. Consequently, when a resident enters one of the sensing areas, the system decides whether he/she belongs to any sensing area by integrating the sensing information collected from all of the PIR sensors in the room. For example, when a resident enters the sensing area B, sensors a and b output ?ON? signals, while sensor c outputs ?OFF? signal. After collecting outputs, the algorithm can infer that the resident belongs to the sensing area B. According to the number of sensors and the arrangement of the sensors signaling ?ON?, the resident?s location is deter-mined in the following manner. First, if only one sensor outputs ?ON? signal, the resident is regarded to be at the center of the sensing area of the corresponding sensor. If the outputs of two adjacent sensors are ?ON?, the resident?s location is assumed to be at the point midway between the two sensors. Finally, if three or more sensors signal ?ON?, the resident is located at the centroid of the centers of the corresponding sensors. For example, it is assumed that the resident is located at point 1 in the figure when only sensor a signals ?ON?, while the resident is located at point 2 when sensors a and b both output ?ON? signals.
The location accuracy of this system can be defined the maximum distance between the estimated points and the resident. For example, when a resident enters sensing area A, the resident is assumed to be at point 1. On the assumption that a resident can be represented by a
5
point and the radius of the sensing area of a PIR sensor is 1 m, we know that the location accuracy is 1 m because the maximum error occurs when the resident is on the boundary of sensing area A. Alternatively, when the resident is in sensing area B, the resident is assumed to be at point 2, and the maximum location error occurs when the resident is actually at point 3. In this case, the error is 3 / 2 m which is the distance between points 2 and 3. Therefore, the location accuracy of the total system shown in Fig. 3 can be regarded as 1 m, which is the maximum value of the location accuracy of each area. Since the number of sensors and the size of their sensing areas determine the location accuracy of the PILAS, it is necessary to arrange the PIR sensors properly to guarantee the specified system accuracy.
Fig. 3. The location-recognition algorithm for PIR sensors.
In order to determine the resident?s location precisely and increase the accuracy of the system, it is desirable to have more sensing areas with given number of sensors and to have sensing areas of similar size. Fig. 4 shows some examples of sensor arrangements and sensing areas. Fig. 4(a) and 4(b) show the arrangements with nine sensors that produce 40 and 21 sensing areas, respectively. The arrangement in Fig. 4(a) is better than Fig. 4(b) in terms if the number of sensing areas. However, the arrangement in Fig. 4(a) has some areas where a resident can not be detected and lower location accuracy than that in Fig. 4(b). Fig. 4(c) shows an arrangement with twelve sensors that five 28 sensing areas without any blind spots.
6
Fig. 4. Location accuracy according to the sensor arrangement of PIR sensors. (a) 40 sensing areas. (b) 21 sensing areas. (c) 28 sensing areas
with twelve sensors.
When PIR sensors are installed around the edge of a room, as shown in Fig. 4(c), it sometimes may give awkward results. One example is shown in Fig. 5. Fig. 5(a) shows the path of a resident. If we mark the estimated points by using the sensor location or the midpoint of adjacent sensors, it will be a zigzagging patterns as shown in Fig. 5(b). In order to alleviate this, we may regard the sensors on the edges to be located a little inwards, which give the result shown in Fig. 5(c).
7
Fig. 5. The effect of compensating for the center point of the outer sensors. (a) Resident?s movement. (b) Before compensating for the outer sensors. (c)
After compensating for the outer sensors.
3. PERFORMANCE EVALUATION OF THE PILAS
3.1 Resident-detection method using PIR sensors
Since the PILAS recognizes the resident?s location by combining outputs from all the sensors belonging to one cell, determining whether a single sensor is ?ON? or ?OFF? directly influences location accuracy. In general, because the ?ON/OFF? values can be determined by comparing a predefined threshold and the digitized sensor output acquired by sampling the analog signal from a PIR sensor, it is necessary to choose an appropriate signal level for the threshold. For example, Smart Floor, which is another non-terminal method, can recognize a resident?s location exactly by comparing the appropriate threshold and a sensor value, because a pressure sensor outputs a constant voltage based on the resident?s weight when he remains at a specific point. However, because a PIR sensor measures the variation in the infrared signal produced by a moving human body, its output is in analog form, as shown in Fig. 6. That is, as the variation in the infrared
8
相关推荐：
loading