W.G. Teich, M. Seidl, "Code Division Multiple Access Communications: Multi-user Detection based on a Recurrent Neural Network Structure", IEEE Trans. Veh. Technol., vol. 46, pp. 979-984, July 1996.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....The first implicit Hopfield type of receiver was introduced by Varanasi and Aazhang [3] They defined a multi stage detection scheme without explicitely mentioning the term neural . Based on these initial results, many authors, e.g. Miyajima, et al. 6] Kechriotis et al. 4] and Teich et al. [7] have shown that this structure is identical to and can be replaced with the recurrent neural network detector, often referred to as Hopfield net detector. However, due to local optimization, this structure does not provide optimal detection. Lots of articles have discussed modified recurrent ....

....VT Y(k) 5) where Y = Y i ] V = V j ] and W = W ij ] respectively. This quadratic form is often referred to as Lyapunov function. One must note the visible similarity between (3) and (5) which makes possible to use the Hopfield net for the purpose of MUD with the following replacements [6] 4] [7] W lj = # lj , if l #= j W ii =0 V k = b k (6) HNN k = lim i## Y k (i) Since the local optimization behaviour of the Hopfield net, this optimization process can get stuck in one of the local minima which can severely deteriorate the performance. In the next section we propose an ....

[Article contains additional citation context not shown here]

W.G. Teich, M. Seidl, "Code Division Multiple Access Communications: Multi-user Detection based on a Recurrent Neural Network Structure", IEEE Trans. Veh. Technol., vol. 46, pp. 979-984, July 1996.

....procedure is decreased [9] Furthermore, using the concept of MMSE feedback, an analytical rule for slope adaptation results, which seems to be well suited to the problem. This is in contrast to related work on the application of neural networks to multiuser detection, e.g. 10] 11] [12], where an activation function with a xed, empirically optimized slope has been applied. V. Numerical Results and Discussion In contrast to conventional DFE, performance of the proposed iterative soft decision feedback algorithm has not been able to be analyzed analytically, yet. Therefore, ....

W.G. Teich and M. Seidl, \Code Division Multiple Access Communications: Multiuser Detection based on a Recurrent Neural Network Structure," in Proceedings of the Fourth International Symposium on Spread Spectrum Techniques and Applications (ISSSTA'96), Mainz, Sept. 1996, pp. 979-984.

....[1] yield poor performance in the case of numerous or differing power users. On the other hand, to implement the optimal Bayesian decision meets severe combinatorial limits. Some authors proposed the use of neural architectures, like Hopfield network, to overcome these difficulties [2] 3] [4] in order to provide an alternative method for optimization. Unfortunately, the Hopfield model can get easily stuck into a local minimum, therefore it may not yield the optimal detection, although its performance is good enough. The authors have developed a brand new detection scheme, named as ....

....to simulate the performance of the optimal detector (the global minima of the associated quadratic form) due to its computational complexity. That is the reason why the theoretical BPSK performance is depicted on the figures that is expected to be very close to the one of the optimal detector [4]. In our complex baseband equivalent model the bit error ratio of BPSK Additive White Gaussian Noise (AWGN) can be calculated as 3 BER = Q s E b N 0 2 = q 2E b N 0 Z # 1 # 2# e x 2 2 dx. We used a five path propagation Rayleigh fading influenced radio environment for ....

W. G. Teich, M. Seidl, "Code Division Multiple Access Communications: Multi-user Detection based on a Recurrent Neural Network Structure" IEEE Trans. Veh. Technol., vol. 46, pp. 979-984, July 1996.

....replaced by soft decision. This was firstly introduced in an indirect way by considering Hopfield neural networks for b b b 1 T b b b 1 2 K n x 1 x 2 x K 2 T K T y 1 2 y K y Detection Algorithm 1 2 K x x x Fig. 1. Modell of code division multiple access communications. multiuser detection [8, 9, 10], which in this case do exactly the same as iterated soft decision interference cancellation (ISDIC) Their good performance may be understood taking into account, that they work in a similar way as the decoder for turbo codes, introduced by [11] Therefore, they can be considered as a kind of ....

....matrix transposition. Then ISDIC is defined by the algorithm for calculating an estimate x for the transmitted vector x shown in Fig. 2. State of the art is using the functions g( Delta) tanh( Delta) and f(R; Delta) d; oe 2 n ; k) c with the constant c being optimized empirically, cf. [8, 9, 10]. Now, a more reasonable choice of the function f( Delta) is shown: The best estimate according to minimum mean squared error (MMSE) in the decision variable d k is the conditional (a posteriori) expectation for x k [12] i.e. d k = Efx k jy k g. If it is assumed, that distortion is only caused ....

W. Teich and M. Seidl. Code division multiple access communication: Multiuser detection based on a recurrent neural network structure. In Proc. of IEEE ISSSTA, pages 979--984, Mainz, Germany, Sep. 1996.

No context found.

Shaker, Aachen. Teich, W. G. and Seidl, M. (1996). Code division multiple access communication: Multiuser detection based on a recurrent neural network structure.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC