Acoustic vector sensor based intensity measurements for passive localization of small aircraft INTRODUCTION

Transcription

1 Acoustic vector sensor based intensity measurements for passive localization of small aircraft S.Sadasivan, Former Scientist G, ADE, Bangalore, India Tom Basten,TNO Science and Industry, PO Box 155, 2600 AD Delft, The Netherlands Hans-Elias de Bree, Microflown Technologies, PO Box 300, 6900 AH Zevenaar, The Netherlands, INTRODUCTION It is well known that propagating acoustic waves from an aircraft can be intercepted by a ground based receiver for gleaning information on relative location and dynamics of the source. It is remarkable that French engineers deployed acoustic arrays as early as World War-I in visionary attempts at remote and passive aircraft detection [1]. Clearly, applications abound given the passive nature of such a localization approach in battlefield situation awareness, border control, helicopter navigation, sniper detection, machinery health monitoring and active noise control. The acoustic waves can be of transient nature as in a gunshot event or short term stationary as in aircraft and land vehicle operation scenarios. The Doppler effect has been significantly exploited for passively estimating trajectory parameters speed, rest source frequency and scalar distance to the closest point of approach to a ground based receiver of an over flying aircraft using a constant velocity straight and level flight model. Essentially this model has been central to the investigation of many real world application scenarios. The focus has been on methods to estimate Doppler frequency variation and solution approaches to the nonlinear estimation problem for source dynamic parameters. An algorithm that estimates a projectile s rest frequency and minimum distance to the receiver in a least squares sense using a single ground based microphone measurements is an early work on the promise of emerging digital signal processing hardware in real-time estimation applications [2]. Use of the Doppler technique for helicopter speed and range prediction based on phase information from successive time interval measurements for frequency estimation is comprehensively discussed in [3]. A similar constant velocity and straight line trajectory model has been used for kinematic parameter estimation of a low flying aircraft using the time variation of the Doppler frequency arrived at employing joint time-frequency distribution methods, again in a single ground based receiver configuration [4,5].

2 The microphone pressure measuring sensor has been the sensor of choice - singly or in an array configuration - in these applications. An entirely new approach that focuses on use of an acoustic vector sensor was first presented by Hawkes and Nehorai with aircraft localization in battlefield environment as the scenario of relevance [6]. The availability of a sophisticated acoustic vector sensor the Microflown has now made possible the realization of 3D intensity based algorithms in passive acoustic modality applications [7]. This sensor measures, in a MEMS configuration, truly the three particle velocity components and together with a conventional microphone enables complete description of the acoustic flow structure at a single spatial point. Note that pairs of pressure microphones are scalar sensors and not true particle velocity sensors and the gradient based velocity estimates provided by such a configuration are not very accurate [8]. A preliminary work using microflowns for helicopter tracking is described in [9]. The use of Doppler shifted acoustic intensity obtained from microflown measurements for aircraft directional localization is brought out in [10]. Presented herein are results of evaluation of the performance of the microflown in real world context using field measurements involving low flying propeller aircraft. In addition to the non-parametric approach, the parametric approach based on co-spectra computed using the Burg maximum entropy method (MEM) is investigated for acoustic intensity estimation towards source directional localization. MEASUREMENTS AND SIGNAL PROCESSING Measurements were made at the Teuge airport, Netherlands. Details of 3D microflown sensor locations with respect to runway and aircraft flight and signal acquisition are provided in [11]. Processing of the vector signals was carried out offline. Both conventional and parametric spectra were used in co-spectra estimation towards computing acoustic intensities. Typically, a length of about one minute of signals was simultaneously captured in the data acquisition process. The block length for Welch method was s and for Burg 0.1 s. For Welch spectra, the simultaneously sampled signals are segmented into 2048 point successive stretches for computation. For each stretch, a 1024 point FFT spectrum is computed with 75% overlap for spectrum averaging. Given the sampling rate is 6000 Hz after four successive decimation by 2 of the signals original sampling rate being Hz, the frequency resolution is Hz. The Burg spectra are computed using 0.1 s stretch lengths( 600 points) and an order of

3 120. The model order was chosen as 20 % of stretch length. The Burg spectra were computed at intervals of Hz (6000/2^13). RESULTS AND DISCUSSION Although measurements were made of six different aircraft transits, only select results are presented herein. Fig.1 Propeller Aircraft in Flight Typical Measurement Scenario Fig.2 Sensor Location and Orientation with respect to Runway Shown in Fig.1 is a typical propeller aircraft a pusher - in straight and level flight at low altitude. The sensors were positioned with axes as shown in Fig.2 with respect to runway. The spectrogram structure of pressure and the three orthogonal particle velocity signals is shown in Fig.3 for a duration of one minute of flight. Note only the band between 200 Hz and 500 Hz is shown with magnitude in log scale as the particle velocity sensor frequency response was flat in this region of frequencies. Harmonic features are evident in the spectra with at least 4 dominant tones possibly at intervals Fig.3 Spectrogram - Pressure and the 3 proportional to propeller rate. Sub harmonics are Orthogonal Particle Velocity Components also seen. The irregular features in the early epochs in spectrogram denote voice of the investigator. The variation in frequencies as a function of time is due to the Doppler effect. The variation is seen similar in pressure as well as particle velocity patterns. Clearly, the single sensor Doppler effect based estimation of aircraft parameters so far carried out using microphone measurements [4,5] can equally well be performed using particle velocity signals. The power of vector sensor in directional localization is realized from joint pressurevelocity correlations, more specifically the acoustic intensity defined as the co-spectrum of these two quantities. Although periodogram based cross spectra estimation is standard, it is attractive to consider parametric approaches to co-spectrum analysis because short signal lengths can be used consistent with stationarity assumptions as regards the moving source. The co-spectrum can be shown to be equal to Fig.4 Horizontal Intensities Top Panel : Welch Method Bottom Panel: Burg MEM Fig.5 Horizontal Angles Top Panel : Welch Method Bottom Panel: Burg MEM

4 half the value obtained by subtracting sum of the auto spectra of the two signals from the auto spectrum of the sum of the two signals [12]. Since this definition entails only auto spectra, one can use parametric methods for acoustic intensity computations. The horizontal acoustic intensity (X and Z ) comparison between the periodogram approach and Burg Maximum Entropy Method shown in Fig.4 brings out how the instant when one intensity reaching the maximum with the other the minimum, is useful in identifying closest point of approach (cpa) as also flight direction as could be inferred from the X intensity variation. Note that the X axis of the sensor is roughly parallel to the runway and hence flight trajectory. That the shorter lengths used in MEM could characterize even the hand sound of the investigator - to denote the closest point of approach of the aircraft to sensor location at about 32 s - is a vindication of the power of parametric spectral approach in the context of moving source tracking requirements. The horizontal angle time history of the source with reference to sensor axes, computed from the horizontal intensities is shown in Fig. 5. The parametric spectra based computations are in agreement with the conventional approach and the results are consistent with physical scenario. Significant azimuth coverage is seen to be provided. The aircraft kinematic parameters were obtained from the Doppler frequency variation obtained from vertical velocity signals, the procedure being same as for pressure signal measurements[4,5]. The harmonic around 220 Hz as computed from Burg MEM was used for estimating the parameters. The results speed 32.6 m/s, distance to closest point of approach (cpa) m and time at cpa s were used for estimation of the angle between aircraft trajectory and radial joining aircraft and microflown sensor location, Fig. 6. A comparison of this estimate with horizontal intensity based values is shown in Fig.7. Excellent correlation is seen in the region of flight where the straight and level model is presumed valid. Fig.6 Idealized Geometry Aircraft Flying by Microflown Station Fig.7 Top Panel Doppler Frequency Variation in Vertical Velocity Signal Bottom Panel Azimuth from Vertical Velocity Doppler and Horizontal Velocity Intensities Given that the horizontal angle is an important information as regards target acquisition and tracking applications in passive surveillance based on acoustic modality, the computation of the angles is relatively straightforward and simple if the vector intensities are used. Note Doppler based approach requires frequency computations and least squares method for parameter estimation and thence the angle. The microflown is thus a strong candidate sensor in the node of an unattended ground sensor

5 (UGS) network optimally exploiting the digital signal processing resources of the node for detection and tracking of air and ground targets. CONCLUSIONS Performance of a modern acoustic vector sensor the microflown has been characterized in real world application scenario concerning low flying propeller aircraft localization. Acoustic intensity based direct estimation of horizontal angle time history for aircraft motion is shown to be consistent with actual flight of the aircraft. Use of parametric co-spectrum in intensity and hence in horizontal angle computations is shown a viable and powerful option. Correlation of horizontal angle estimates from Doppler frequency time history of the vertical particle velocity signal and from horizontal intensities is shown to be very satisfactory and brings out computational trade offs. The advantage of direct computation of horizontal angles from acoustic intensities afforded by microflown measurements, is relevant to development of vector sensor localization algorithms in unattended ground sensor network deployment for acoustic modality based passive and persistent surveillance requirements. ACKNOWLEDGEMENT The first author is grateful to Mr. Alex Koers, Director, Microflown Technologies for providing internship support at the HAN University Campus, Netherlands. Students Michael Maassen, Jordy de Boer and Lucas Haude are thanked for their valuable assistance in measurement and analysis. REFERENCES 1. Johnson, D.H., and Dudgeon,D.E., Array Signal Processing: Concepts and Techniques,Prentice Hall, Englewood Cliffs,NJ, Statman, JI., and Rodemich, ER., Parameter Estimation Based on Doppler Frequency Shifts, IEEE Transactions on Aerospace and Electronic Systems, Vol.AES 23, No.1, January 1987, pp Olson, R., and Cress D.H., Passive Acoustic Range Estimation of Helicopters, Technical Report EL-92-13, Environmental Laboratory, Department of the

6 Army, Waterways Experiment Station, Corps of Engineers, Vicksburg, Mississippi, Final Report, February, Ferguson, B.G., and Quinn, B.G., Application of the Short-Time Fourier Transform and the Wigner-Ville Distribution to the Acoustic Localization of Aircrfat, Journal of Acoustical Society of America, Vol.96, 1994, pp Sadasivan, S., et al., Acoustic Signature of an Unmanned Air Vehicle Exploitation for Aircraft Localisation and Parameter Estimation, Defence Science Journal, Vol.51, No.3, July 2001, pp Hawkes, M., and Nehorai, Arye., Wideband Source Localization Using a Distributed Acoustic Vector-Sensor Array, IEEE Transactions on Signal Processing, Vol.51, No.6, June 2003, pp H.E. de Bree, P. Leussink, T. Korthorst, H. Jansen, T. Lammerink, M. Elwenspoek, The Microflown: a novel device measuring acoustical flows, Sensors and Actuators A-Physical, 54, pp , (1996). 8. Nehorai, Arye., and Paldi, Eytan., Acoustic Vector-Sensor Array Processing, IEEE Transactions on Signal Processing, Vol.42, No.9, Sept. 1994, pp T. Basten, H.E. de Bree and E. Tijs, Localization and tracking of aircraft with ground based 3D sound probes, European Rotorcraft Forum 33, Kazan, Russia, (2007) 10. S. Sadasivan, et al., Sensors and Signal Processing in Passive Acoustic Modality Estimation of Aircraft Flight Dynamic Parameters", Proceedings of the International Conference on Avionic Systems, ICAS , pp Tom Basten, Hans-Elias de Bree, S. Sadasivan., Acoustic eyes, a novel sound source localization and monitoring technique with 3D sound probes, accepted for presentation at ISMA 2008 International Conference on Noise and Vibration Engineering, September 15-17, 2008, Leuven, Belgium 12. TM Romberg, JL Black and TJ Ledwidge, Signal Processing for Industrial Diagnostics, John Wiley & Sons, 1996