The use of optical sensing technology to extract vibration sources has been successfully demonstrated in previous works. In a system described in Z. Zalevsky et al., an optical system was designed and demonstrated for the detection of sounds by projecting a laser beam on a vibrating target and observing the movement of the secondary speckle patterns that are created by the interferences from the target’s roughness [1
]. The speckles are self-interference random patterns. It has the unique quality that each individual speckle serves as a reference point. From each speckle point, one may track the changes in the light phase that are being scattered from the target surface. This speckle-pattern interferometry was used for measuring displacements and analyzing vibration frequencies, as well as characteristics of the object deformations. In order to measure the object deformation displacements, the speckle pattern is subtracted before and after the deformation has occurred. The deformations are due to changes in loading, temperature, etc. This procedure produces a speckle pattern that corresponds to the object’s local surface changes between two exposures. From the correlation between the speckle patterns, both the magnitude and the direction of the object’s local surface displacement are able to be determined.
Other applications which utilize an optical approach for the extraction and separation of remote vibration sources were used for biomedical measurements, such as monitoring heart rate, breathing, blood pressure, blood oximetry, blood coagulation [1
], bone fractures [5
], melanoma [6
], and glucose concentration in blood [7
The above applications illuminate the object under observation using a laser beam which produces the secondary speckle pattern. A fast-imaging camera observes the temporal intensity fluctuations of the imaged speckle pattern and its relative movements [9
]. In this study, we show that the same algorithm used for optical sensing can be used also when the object under measurements is excited by an RF source. Hence, instead of having the image recorded by a camera, the scattering is received and processed using an RF receiver.
In the following experiment, we used RF frequency pulses for generating the secondary speckle patterns. The illuminated object was an automotive vehicle. It was illuminated under two different conditions. The first was when the vehicle engine was turned on, and the second was with the engine turned off.
A series of pulses were transmitted toward the vehicle. The returning pulses were stored in a range bin (discrete data elements received from the reflected radar signal) and processed by applying an image correlation algorithm. We found out that not only was it possible to detect the movement of the vehicle, but it was also possible to determine its vibration frequency.
The use of RF pulses, as opposed to optical pulses, enables us to determine the vibrations of an observed object at far greater distances than what would be possible in an optical system.
In our experiment, a transmitter was utilized and operated at 10 GHz. Increasing the transmitter frequency increases the sensitivity resolution of the entire system. This would enable us to detect vibrations at higher frequencies. However, doing so in an optical system would require using a complicated integration and expensive hardware.
In addition, the use of a Doppler radar enables us to detect both the vibration frequencies and the directional velocity by increasing the sampling rate. As opposed to optics, the radar field-of-view is relatively wide, and can operate without a tracking system as used in optical systems [1
Several options were considered for illuminating the vehicle. The preferred option for our experiment was using an X-Band, phased array, Pulse-Doppler radar. This radar is normally used for search and detection of personal in the surrounding areas. Its main use is for detection of threats, threat range, and threat velocity. Each radar section has a transmitter (Tx) and receiver (Rx) modules connected directly to the antenna.
The antenna is a phased array antenna consisting of 16 radiating elements. Normally, the 16 antenna beams are combined into a single beam which scans in the horizontal (azimuth) direction. In our demonstration, the scanning option was disabled, and only eight channels were operated simultaneously.
The vertical (elevation) 3 dB beam width was 10°, while the vertical (azimuth) 3 dB beam width was about 100°. The beams were separated by 0.5° and the pulse width was 0.1 ms with a dwell time (illumination time on target) of 100 ms. The tested vehicle was illuminated at ranges varying from 50 m to 120 m.
Note, that even though phased array radar was used in our experiment, many other RF transmitting techniques are also valid.
2. Experimental Setup—Pulse-Doppler Radar
The setup was consisted of a Pulse-Doppler radar and a vehicle. The measurement was done in an open field at Bar-Ilan University. The radar was placed at one end of the field, while the vehicle (Toyota Corolla Sun, Japan, 2008) was placed at various distances from the radar in front of the radar line-of-sight. The radar system is shown in Figure 1
The distance between the radar and the vehicle was set to 40, 80, and 120 m, respectively. As was mentioned earlier, the vehicle was illuminated by the radar in two different scenarios. In one scenario, the engine was off, while in the second, the engine was turned on. The measurements of the received pulses were processed using a unique algorithm designed to extract the secondary speckles reflected from the vehicle, and to determine its rate of vibration when the engine was on. Prior to each measurement, a person walked at the targeted site, which was next to the vehicle. This was done in order to provide a reference for detecting the vibration of the vehicle. The discrimination between the movements of a person and a stationary vehicle is possible, since the signal returning from the target was measured using a Pulse-Doppler radar, in which the strength of the received pulse was a function of the target’s movement with respect to the radar.
The principle behind the Pulse-Doppler radar is that it is used to calculate the range to targets by measuring the elapsed time between sending a radio pulse and receiving the reflection from the object [14
]. It also utilizes the Doppler effect, where the target’s movement produces a frequency shift on the signal reflected back from the moving target. As the target moves between each transmitted pulse, the returned signal has a phase difference or phase shift from pulse to pulse. Since the movement velocity of a person in reference to the radar is higher than that of the vibrating stationary vehicle, the amplitude of its Doppler frequency will be higher than that of the vehicle, thus enabling it to be used as a relative reference signal.
The goal of the pulsed-Doppler signal processing is to extract the desired target signal out of the surrounding noise or the clutter (unwanted echoes). It is done by using techniques which allow small high-speed objects to be detected in close proximity to large and slow-moving objects. It organizes the measured samples into an m by n matrix. Figure 2
is an illustration of such a matrix. The m by n data cubes in the time domain is shown in the top half of the diagram. Each data sample in the “range samples” axis represents the distance from the radar antenna.
When the radar illuminates a target, multiple returns are received and processed. The “pulse interval” axis represents the range to the target extracted from the returning radar pulses. Each column in the “pulse interval” axis represents the individual samples taken between each transmit pulse [18
]. There is an individual cube for each pulse repetition interval (PR). Since the Pulse-Doppler radar measures the Doppler shift and the fact that the target is moving radially, there will be a time shift between each pulse, which results in a phase shift over multiple returns. The time samples are converted to the frequency domain using a digital processor. This involves the Fast Fourier Transform (FFT) algorithm.