Neural Network A SHORT INTRODUCTION TO NEURAL

16 The Performance of Supervised and Unsupervised Neural Networks in Performing Aircraft Identification Tasks Arwin Datumaya Wahyudi Sumari 1 Aciek Ida Wuryandari 2 Maman Darusman 2 Nur Ichsan Utama 2 1 Departemen Elektronika, Akademi Angkatan Udara Jl. Laksda Adisutjipto, Yogyakarta 55002 – INDONESIA Telp. 0274 486922 ext 6101 Fax. 0274 488918 Email: arwin91aauyahoo.co.id 2 Sekolah Teknik Elektro dan Informatika - ITB, Kampus ITB Labtek VIII Lantai 2, Jl. Ganesa 10, Bandung 40132 – INDONESIA Telp. 022 2502260 Fax. 022 2534222 Email: acieklskk.ee.itb.ac.id , maman_darusmanyahoo.com , nur_ichsanymail.com ABSTRACT This paper is a report on our research progress in the area of aircraft identification by utilizing neural networks and information fusion. In this paper we address the performance comparison of supervised and unsupervised neural networks in aircraft identification tasks in a generic system called Neural Network-based Aircraft Identification System NN-AIS. We select Adaptive Resonance Theory ART for the unsupervised neural network and Back Propagation Network BPN for the supervised one. As for previous research, we use two kinds of input namely aircraft Radar Cross Section RCS and average speed. Their performance will be validated by using already-learnt and never-learnt patterns . Keywords : ART, BPN, aircraft identification, RCS, speed

1. INTRODUCTION

Aircraft identification task is a critical matter to recognize the identity of an aircraft that is entering a monitored air space. The sooner the observed aircraft is identified, the faster the authorized authority can make a decision. In normal flight procedure, all aircraft flight plans must be reported to the authorized authority to be recorded. The records will be used to monitor every single aircraft movement in the monitored air space. The reported flight plans ease the authority to track and identify a certain aircraft that is displayed on monitor room’s displays. A problem is arisen when the authority cannot identify a certain aircraft that is detected by radar system. There are two possibilities when an aircraft cannot be identified. First, there is a possibility the aircraft transponder for answering the interrogation signal from ground station is not working properly or failed. Second, there is an intention to turn-off the transponder in order to hide the aircraft identity. For the second reason, we can conclude that the aircraft must be in undercover missions and can be a threat to our country’s sovereignty. Because of it, the authority must have a way to identify the unidentified aircraft before something harmful occurs in the future. One of ways in identifying aircraft is by using its Radar Cross Section RCS value and combining it with its speed. On the other hand, one approach that has been known well for object recognition is neural networks. In this paper we address the utilization of two kinds of neural network architecture for performing aircraft identification tasks namely supervised and unsupervised. We use the two types of input for training and validating the networks and measure their performance. In general our proposed system is called as Neural Network-based Aircraft Identification System NN-AIS. The structure of the paper is as follows. Section 1 cover the background of the paper and it will be followed by Section 2 which covers a short introduction to neural networks as well as related matters to aircraft RCS and speed. Section 3 presents the NN-AIS design as well as the NN training. Section 4 delivers the system validation. The paper is summarized in Section 5 with some concluding remarks.

2. A SHORT INTRODUCTION TO NEURAL

NETWORK AND AIRCRAFT RCS In this section, we will present a short introduction to neural network, its learning paradigm taxonomy, and a brief explanation regarding two types of learning paradigm we select for our research. We also deliver a very brief explanation regarding aircraft RCS and speed.

2.1. Neural Network

The most basic constructing element of a human nervous system is a neuron which is called as “processing unit” as presented in Figure 1. According to Shepherd and Koch 1990 in Haykin 1994, human brain has more than 10 billion neurons and 60 trillion synapses or connections between neurons. Even though it is relatively slower than computer systems that are made up from nano-technology silicon gates, it can do highly complex, nonlinear, and parallel tasks such as pattern recognition and perception, faster and very much better than the best computing system that human ever created. 17 Figure 1: Neuron or nerve cell. Because the neural networks are good for recognition tasks, some earlier researchers such as McCulloch-Pitts, Grossberg, Minsky, etc., tried to model the nervous processing unit so its mechanism can be emulated in computing systems. From this perspective, we define Artificial Neural network ANN or usually called as just Neural Network NN as an emulation of human nervous system when performing information processing. Its characteristics are displayed on the ability to obtain new knowledge after a successful learning process and store it in the information storage which is its synaptic weights. In more detail, NN is generalization of mathematical models of human cognition based on the assumptions that Fausset, 1994: • information processing occurs at many simple elements called neuron, • signals are passed between neurons over connection links, • each connection link has an associated connection weight which multiplies the signals transmitted, • each neuron applies an activation function which is usually non linier, to its net input to determine its output signals. Figure 2 : A mathematical model of a neuron. In NN model, neuron takes a set of inputs, , along with a set of connection or link or synaptic which are characterized by weights, . The summing junction, ∑, sums up the input signals that are amplified by the connection weights. The activation function, m x km w . ϕ , limits the net outputs in allowable values. The architecture of the NN model is depicted in Figure 2. The general mathematical equations for neural information processing are given in Equation 1 for inputs summing process to obtain k v = = ∑ m k kj j v w j x 1 and Equation 2 for producing the NN output, . k y k k y v ϕ = 2 2.2. Neural Network Learning Model Taxonomy According to Haykin’s 1994 taxonomy, there are three NN learning models. • Supervised. The essential of this paradigm is the availability of an external supervisor, so there will be an input-output relation in order to find the most minimum disagreement between the NN outputs with the examples given by the supervisor. • Unsupervised or Self-Organized. In this learning paradigm, there is no external teacher or examples to be learnt by the NNs. So, the NNs will perform a competitive learning rule where the winning neuron is entitled to keep the input in its memory. • Reinforcement Learning. This is the on-line learning of an input-output mapping through a process of trial and error designed to maximize a scalar performance index called as reinforcement signal.

2.3. Supervised Neural Network – Back