A Novel Technique or Blind Bandwidth Estimation of the Radio Communication Signal

International Journal of ISSN 0974-2107 Systems and Technologies IJST Vol.3, No.1, pp 11-16 KLEF 2010 A Novel Technique or Blind Bandwidth Estimation of the Radio Communication Signal Gaurav Lohiya 1, P. S. Prasad 1, G. Raghavaiah 1 1 Defence Electronics Research Laboratory, Hyderabad, India gauravlohia@dlrl.drdo.in Abstract: This paper describes a computationally efficient technique for blind bandwidth estimation of the radio communication signal. The technique is based on applying statistical techniques to a power spectrum of the received signal. The performance is illustrated on the blind bandwidth estimation of the simulated radio communication signal. Applications include using the estimated bandwidth to reduce the false alarm rate of signal detection and setting the appropriate bandwidth in the receiver to get the advantage of SNR improvement, thereby making the further signal processing more efficient. 1. INTRODUCTION This paper describes a computationally efficient technique for blind bandwidth estimation of the radio communication signal. The technique is based on applying statistical techniques to a power spectrum of the received signal. This developed technique does not require a- priori knowledge about any characteristics of the signal parameters such as: frequency, SNR, type of modulation and parameters of modulation. Moreover, this technique does not assumes any noise model and works efficiently even for the signals having low SNR of around 6 db. Estimating the bandwidth with statistical techniques is motivated by the fact that the humans are good at estimating the bandwidth by eyeballing a spectral plot. Intuitively, we Figure 1: Signal spectrum 11

Gaurav Lohiya separate the spectral humps from the noise floor by eliminating those parts of the spectrum shape that are due to signals and visually draw in the noise floor. Accurate bandwidth estimation is important in a number of signal detection problems. Figure 1 shows an example of a simulated signal spectrum. Blind bandwidth estimation of signals of unknown carrier frequency in a frequency band of interest much wider than the signal of interest is a common problem in many fields. A common bandwidth estimation technique is based on measuring the power spectrum over the bandwidth of interest and comparing the received signal power spectrum with the noise floor plus a threshold. The threshold level affects the accuracy of correct bandwidth estimation and probability of missed detection. The performance of this approach generally requires an accurate estimate of the receiving system noise floor. One way of establishing the noise floor is to measure it when the signals are not present. This generally requires taking the receiving system off-line and performing lengthy and tedious calibration procedures. There are many reasons that motivate blind bandwidth estimation when signals are present and the system is on-line including: Taking the system off-line may not be an option because other activities require use of the sensor. Oftentimes, estimating the bandwidth requires performing a lengthy and tedious calibration procedure. The noise floor may change with time due to component aging as well as environmental conditions such as temperature and moisture, which would require frequent calibration. The signal environment may vary the noise floor. This paper consists of three sections: Section 2 provides a background on the problem area and a brief review of Existing techniques; Section 3 describes the developed technique for blind bandwidth estimation; Section 4 is the Performance Example, which, illustrates the blind bandwidth estimation performance. 2. BACKGROUND 2.1 ELECTRONIC WARFARE A definition of electronic warfare (EW) is [1]: z Military action to exploit the electromagnetic spectrum which encompasses the interception and identification of electromagnetic emissions; the employment of electromagnetic energy, including directed energy, to reduce or prevent the 12

A Novel Technique hostile use of the electromagnetic spectrum and actions to ensure its effective use by friendly forces. Radio communication signals may be intercepted and analysed in terms of intensity, direction and type for counter measures such as jamming, or be demodulated for the purpose of intelligence and eavesdropping. In the latter case, determining the bandwidth of the radio signal is crucial. The accuracy of the bandwidth estimate directly affects the performance of the algorithms for intelligence extraction. But also for setting the appropriate bandwidth for the jammer, knowledge of the signal bandwidth is highly beneficial. In a non-co-operative setting, bandwidth estimation introduces a number of challenging requirements. Factors such as noise, interference, propagation effects and spread spectrum techniques may seriously complicate the task, especially if little or no a- priori knowledge is at hand. Determining bandwidth under such conditions usually requires trained operators and signal analysis tools. 2.2 EXISTING TECHNIQUES In this section, background is given on relevant research already existing in the literature. Bandwidth estimation technique operates on spectral amplitude data that has been logarithmically compressed; the outputs of the linear filter processes are related to the geometric mean. The technique is assessed based on false-alarm rates for a given probability of detection and input Signal-to-Noise-Ratio (SNR). This technique uses references [2], [3] & [4]. Adjustable Bandwidth Concept (ABC) approach [5] approach, the spectrogram, a timefrequency representation, is accomplished using the short term Fast Fourier Transform (FFT). The log-spectrogram is formed by logarithmic compression of the magnitude of the coefficients of each of the FFT frequency bins. The time segments can be nonoverlapping for computational efficiency, or overlapped for enhanced time resolution. Likewise, various data windows can be used to control spectral leakage. The resulting n Power Spectral Density (PSD) estimates are averaged. The threshold is set based on knowledge or estimation of the noise floor. (For example, the Rank-Select-Threshold (RST) method of noise floor estimation can be used as a method of automated noise floor estimation. In this method, the content of the PSD bins are sorted in ascending strength, and the strength of the bin at around the p th percentile is selected as the noise floor estimate, where p is some appropriate fraction.) Technique [6] includes generating filter coefficients from a generating value, Obtaining a power measure of a received signal with respect to a selected 13

Gaurav Lohiya Frequency and estimating a bandwidth of the received signal based on the power measure. In obtaining the power measure, a method includes multiplying samples of the received signal with the filter coefficients. Generating filter coefficients includes rotating a filter coefficient by the generating value. This technique includes obtaining power measures of a received signal, each corresponding to a selected frequency, and estimating a bandwidth of the received signal based on the power measures. Non-uniform sampling a frequency spectrum of a received signal is carried out for determining power measures of the received signal at the sampling frequencies, and obtaining an estimate of the bandwidth of the signal based on the power measures. Non-uniformly sampling the frequency spectrum includes filtering the received signal with filters centered at the selected sampling frequencies. A filter includes a lookup table to store generating values, a first multiplier to receive a generating value and a current filter coefficient and to output a subsequent filter coefficient, an accumulator to store the subsequent filter coefficient, a second multiplier to multiply the current filter coefficient with a sample of a received signal and to output a current filtered value, and an adder to receive the current and past filtered values and to output an accumulation signal. A filter includes a power calculator to output a power measure based on a value of the accumulation signal. A system includes a lookup table to store generating values, filters to output power measures, and a bandwidth estimator to output an estimate of the bandwidth of a received signal. The bandwidth estimator compares a relation between two or more power measures to a predetermined threshold and/or modifies one or more power measures based on a power measure corresponding to a frequency outside an expected bandwidth of the received signal. 3. DEVELOPED TECHNIQUE FOR BLIND BANDWIDTH ESTIMATION The developed algorithm consists of estimating the power spectral density of the signal (shown in figure 1), followed by integrating the power spectral density over the frequency band of interest. This integrated spectrum is divided into number of short frames of equal size. The variance is estimated on each frame and the frequency at which the variance is above threshold, is recorded. These frequency points represent the start frequency and stop frequency for spectrum of each signal. The difference between the start and stop frequency is the estimated bandwidth of the signal. In the case of multi-tone signals, like Frequency Shift Keying, Minimum Shift Keying etc., the frequency difference between different tones cannot be more than the guard band between two channels. Thus, the frequencies very close to each other are assumed to be the different tones of the same signal. Hence, the difference between the extreme edges is taken as a bandwidth estimate of that signal. 14

A Novel Technique IV. PERFORMANCE EXAMPLE The bandwidth estimation technique performance is illustrated on a signal received. The signal received is shown in figure 1. The integrated spectrum is shown in figure 2, along with the estimated start and stop frequency points as a circle. From the figure 2 it is clear that whenever there is a sharp transition in the signal, the variance changes and crosses the threshold. These frequency points are mapped onto the signal spectrum to estimate the bandwidth of the signal and are shown in figure 3. Figure 2: Integrated signal spectrum. Circles show the estimated start & stop frequency points in the spectrum The estimated bandwidth matches the actual estimate that would be produced by eyeballing a spectral plot very well. In addition, the signals have only 6 10 db SNR. Straight-line threshold on this signal would cause wrong bandwidth estimates. Figure 3: Signal spectrum with the detected signals and their bandwidths (black line) 15

Gaurav Lohiya 5. CONCLUSION In this paper, we have considered the challenges posed by the EW scenario to existing techniques for bandwidth estimation. We have specifically focused on the development of algorithm, which is independent of the frequency band of operation, frequency of signal etc. We have evaluated the algorithm to handle unknown signals and found it to be adequate. The presented technique determines the bandwidth much the way the eye does with spectral plots by separating shapes. It has been used successfully for a number of real scenarios resulting in high probability of signal detection & bandwidth estimation, even for signals with SNR as low as 6 db. It is more efficient in terms of performance and computations as compared to the existing techniques. Moreover, the presented technique mimics visual intuition. ACKNOWLEDGEMENT The authors are very thankful to the Division Head and Director, Defence Electronics Research Laboratory, Hyderabad for their encouragement and support that helped to improve the paper. The authors also thanks to reviewers for discussions and help in preparing the final version of this paper. REFERENCES Smart Sensor Solution Division: Communications Electronic Warfare [online] (2004): NTO Physics and Electronics Laboratory, The Netherlands. HTTP format. [Cited 8 th July 2004]. Available from <http://www.tno.nl/instit/fel/div3/feld33-2.html>. Rik Pintelon, et. al., The Geometric Mean of Power (Amplitude) Spectra has a Much Smaller Bias than the Classical Arithmetic (RMS) Averaging, IEEE Transactions on Instrumentation and Measurement, Vol. 37, No. 2, June 1988; G. Corsini, et. al., Cramer-Rao Bounds and Estimation of the Parameters of the Gumbel Distribution, IEEE Transactions on Aerospace and Electronic Systems, Vol. 31, No. 3, July 1995; Filippo Attivissimo, et. al., A Study on Nonlinear Averagings to Perform the Characterization of Power Spectral Density Estimation Algorithms, IEEE Transactions on Instrumentation and Measurement, Vol. 49, No. 5, October 2000. Andrew J. Noga, Adjustable bandwidth concept performance evaluation, AFRL-IF- RS-TR-2003-184, In-House Interim Report, July 2003 Patel, Shimman, Wilborn, Thomas B., Method and apparatus for bandwidth estimation, United States Patent 7039138 16