GSM mobile phone based automobile security system

Y.S. Moon, K. Wong and K.S. Ho

A prototype of a novel automobile security system is described. Besides tracking the positions of the target, video and audio signals inside the automobile are also captured. The data are sent to the server through the GSM network for analysis. Experimental results show that the approach is feasible given the low bandwidth of the GSM network.

Introduction: Distant delivery is insecure since what happens dur- ing delivery is often unknown. An obvious solution is to keep track of the object 11, 21 by using the GPS [3]. However some- times, knowledge about the location alone is not sufficient for determining what goes wrong if the object does behave abnor- mally. We may even be unaware of the problem especially if the object is still moving along the predicted path. In view of this, we propose a novel tracking system for automobiles. The system con- sists of two parts: the client and the server. The client is a Pen- tium-233 MMX notebook computer installed in the automobile. It is equipped with a Gannin 45 GPS receiver to collect the posi- tional information of the target. In addition, video and audio sig- nals are gathered inside the automobile using a digital camera and a microphone, respectively. Signals are sampled periodically, which are sent to the server via a GSM mobile phone after com- pression. The (stationary) server then makes use of the GPS sig- nals to plot the path of the target on an electronic map. At the same time, the video and audio signals are also reproduced (Fig. I).

Fig. 1 Server in operation

Znterpolathzg GPS dum: Sometimes, the server may fail to deter- mine the exact location of the target (e.g. signals can be blocked by tall buildings). In such cases, the server has to estimate its path. Consider the following example. Suppose the latest position before the GPS signals become unavailable is Po (Fig. 2). Two consecu- tive check points W, and Wk+, (which are user-specified points

W,+1

-._ \bw. r . _ _ , _ . _ . -.-.;.o _ , _ , -._

p, . ' @33i?j

Fig. 2 Interpolation of GPS data to reconstruct missing path

0 check point 0 GPS data

result of interpolation - _ _ _ actual path received

ELECTRONICS LETTERS 2nd March 2000 Vol. 36

along the path that the target is expected to follow) that are clos- est to PO are located (the line joining them has the shortest perpen- dicular distance to Po). Similarly, when GPS data are available again, the server can use the newly received positional data Pi to locate another pair of check points W, and W,,,. These points can then be used to interpolate the missing path.

Adaptive sampling of GPS data: If data are sampled too often, we may gather a group of closely located GPS data, thus wasting bandwidth. However, if the sampling rate is too low, the sampled GPS data will be widely separated. This will result in a non- smooth path. Regarding this issue, an adaptive method is adopted in which the sampling rate is adjusted dynamically based on the speed of the target. Precisely, given the fact that the path is pre- defined, we can estimate the next position of the target based on its current position, its previous position, and the sampling fre- quency. When the GPS data are received, the actual position is compared to the estimated position. If the difference is within a certain threshold, it means that the target is moving normally as scheduled. We thus adjust the sampling frequency as follows:

W f t = fo - -P

2 whereh is the sampling frequency when the target is first observed to be moving normally andJ; is the sampling frequency at the next time-step t (so, the target has been moving normally during the period from time = 0 to t-1). Effectively, the sampling rate is decreasing gradually. Also, as the period during which the target is moving normally is prolonged, the rate of decrease will be increas- ing at a constant rate w. In fact, J; can be derived from the sam- pling frequency of the last period (i.e. A.,) as follows:

f t = ft-1 + W ' (t + 1) In contrast, if the distance between these two positions is

greater than the threshold, the sampling rate will be doubled such that positional data will be sampled more frequently.

Data compression and transmission delay: The GSM network sup- ports a data rate of 9600 bit/s only [4], which is by no means ade- quate for transmitting the huge volume of multimedia data involved. Hence, the Intel Indeo(R) Video R3.2 codec is employed to compress a video frame of 19k bytes to -Ik bytes. For audio signals, the Microsoft Network Audio codec is employed to give a data rate of 1 kbit/s (8kHz, mono, 8200 baud). The system allows the user to choose between streaming video, streaming audio or high resolution still image. After compression, their data lengths are 816 bytes, 1134 bytes and 19200 bytes, respectively, and their transmissions require 0.76s, 0.95 s and 17.6s correspondingly. For GPS data, the data length is 16bytes which requires 0.02s for transmission.

Experiments: The whole system was tested several times with a vehicle travelling along a pre-defined test path. All data could be reliably received from the client and the missing portions of the path could be interpolated successfully. The line was occasionally cut when the vehicle moved pass cells of the mobile network. Yet the system was able to automatically re-establish connection in those situations.

Conclusion: We have demonstrated the feasibility of using a low bandwidth GSM network for transmitting multimedia and GPS data in building a real time automobile security system. By means of data compression and adaptive sampling, we have successfully reduced the amount of data that needs to be transmitted, thus saving air-time cost.

0 IEE 2000 Electronics Letters Online No: 20000358 DOI: 10. 1049/el:20000358 Y.S. Moon (Department of Computer Science and Engineering, Chinese University of Hong Kong. Shatin, New Territories, Hong Kong, People's Republic of China) K. Wong (Department of Engineering, Cambridge University, Cumbridge, United Kingdom) K.S. Ho (Department of Computing, Hong Kong Polytechnic University, Hung Hom, Kowloon, Hong Kong, People's Republic of China)

No. 5 463

23 December 1999

References

1 LOGSDON, T.: Understanding the NAVSTAR, GPS, GIS and IVHS (Van Nostrand Reinhold, 1995)

2 PAPADOGLOU, N., and sTIPIDis, E.: Short message service link for automatic vehicle location reporting, E1ectro.s. Lett., 1999, 35, pp. 876877 HOFMANN-WELLENOF, B., LICHTENEGGER, H., and COLLINS, J.: GPS theory and practice (Springer-Verlag, 1994)

(Addison-Wesley, 1998)

3

4 GOODMAN, D.J.: Wireless personal communications systems

-30

-40

-50 m U

-60

-70

Ray tracing method for propagation models in wireless communication systems

Zhijun Zhang, Z. Yun and M.F. Iskarider

A new triangular-grid ray tracing (TGRT) method is proposed. It is based on dividing the propagation region into triangles. The number of triangles is decided by the number of vertices of the structures, instead of their dimensions. For the 2D case, it can decreases the visible decision calculation to two edges for each triangle. It is shown that this method unifies the indoor andor outdoor propagation problems and provides savings in computational time.

-

-

-

-

-

Introduction: Propagation prediction is very important in the design of wireless communications systems. The ray tracing method has become a vital tool for propagatilsn prediction. This is especially true for the micro- and pico-cell cases since site specific propagation information is needed in developing these communi- cation systems [l].

The conventional ray tracing method is based on a ray-launch- ing and bouncing procedure which can be very inefficient if no speed-up algorithm is employed. Several schemes have been devel- oped to accelerate the ray tracing proceclure, e.g. the image method, the bounding box method and the utilisation of visibility [2, 31. Although these methods have their own advantages, a more efficient method is needed to cope with the complex and often computationally demanding indoor or indoor/outdoor situations while maintaining good accuracy of the propagation prediction results. In this Letter we propose an efficient ray tracing method based on a triangular division of the propagation space.

TGRT method: The developed triangular-grid ray tracing (TGRT) method can be used in both 2D and 3D problems. In 2D cases, a triangular mesh is used to discretise the propagation region of interest. In 3D cases, tetrahedral or triangular cylinders may be used. We focus on the 2D case; an extension of the procedure for the 3D case is presently being developed.

Fig. I Triangular mesh layout or propagation domain

. . . . . . . . . . . additional grid edges walls

The construction of the triangular mesh is rather straightfor- ward. The entire region of interest is enclosed by a polygon (in our case, a rectangle, see Fig. 1). Each room or building is represented by an edge defined by two points (vertices.). A set of vertices is thus obtained and a triangular mesh is constructed using these vertices and satisfying the condition that 1:ach edge, except the

edges of the bounding polygons, is used by only two triangles. Using this procedure, the total number of triangles, Nlrinngie and the total number of edges, Ne+ can be uniquely determined by the number of boundary vertices, Nby, and the number of all verti- ces N, as follows:

Ntriangle = 2(Nv - 1) - Nbv (1)

Eqns. 1 and 2 show the number of triangles and edges, and hence the proposed solution procedure, depends only on the number of vertices. Since it does not include a search procedure, it provides significant computational advantage.

Fig. 1 shows a triangular meshed layout for a rectangular prop- agation region of interest. This region includes two buildings (defmed by walls), one rectangular and one triangular in shape. There are two types of edge, one is a true wall of a structure in the propagation domain, while the other is an additional grid edge. If a ray crosses a grid edge, such as point 1 in Fig. 1, it just passes by the edge and proceeds to the next one. If a ray hits a wall, such as point 2 in Fig. 1, a reflected andor transmitted ray will be cre- ated according to the electrical properties of the wall.

It should be noted that in each triangle and with the arrival of the propagating ray to one of the edges, only two other edges need to be checked to determine which one will be hit next by the prop- agating ray. This decision can be made fairly easily by observing the sign of the vector multiplication between the rays direction vector and vector from the two points defining the vertices of the edge under consideration (e.g. points 1 and 3 in Fig. 1). Because all the information about the adjacent triangles related to each edge is known, the ray can quickly go through from one triangle to another by a pointer-locating method instead of requiring time consuming searching algorithms.

... I:. ~ ... .. . . .

. .. . .. . . . . .- : :: . .

0 40 80 120 160 200 a b

1401/21 Fig. 2 Triangular mesh layout for indoor-outdoor propagation domain and detailed triangular meshing in building 4 a Triangular meshed layout for indoor-outdoor propagation domain b Detailed triangular meshing in building 4 __ walls

additional grid edges

I 1 1, 200 300 400 500 600

time delay, ns 1401/31

Fig. 3 Time delay spread result for indoor-outdoor propagation in building 4

464 ELECTRONICS LETTERS 2nd March 2000 Vol. 36 No. 5