Biomimetic photonics extract the good design of nature and mimic it with photonics. The weakly electric fish genus, Eigenmannia, has a unique neural algorithm – jamming avoidance response, to facilitate their survival in the deep dark ocean, by automatically adjusts the local transmitter carrier frequency to move away from the jamming frequency when it is within the jamming spectral range. Examining our own wireless microwave systems, the situation of inadvertent jamming is very similar as that in Eigenmannia. In this article, a biomimetic photonic approach inspired by the jamming avoidance response in a weakly electric fish genus, Eigenmannia, is naturally adopted to experimentally tackle signal jamming in wireless systems. Mimicking the system with photonics enables the proposed scheme to work for frequencies from hundreds of MHz to tens of GHz.
© 2018 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
Neural algorithms and biological characteristics in animals and plants are being optimized for survival and are extremely efficient in performing the designed tasks. There is lots of hidden treasure in our nature that could be a natural solution towards the critical challenges that we are facing in modern technologies. Discovering those hidden treasure, understanding them, and using photonics to mimic the useful neural algorithm is a new and exciting field. Here, we will study the neural algorithm - jamming avoidance response in a weakly electric fish genus, Eigenmannia – and assemble it with photonics to tackle a long lasting challenge in wireless microwave systems.
Wireless microwave technology plays an important role in human civilization, providing improvements in quality of life, and enhancing national security. The wireless nature of microwave systems enables the birth of a large variety of important technologies. Unfortunately, the wireless transmission media and the broadcast nature of most microwave technologies make them particularly susceptible to jamming [1,2]. Intentional jamming is usually from a noise source, while inadvertent jamming typically is a modulated sinusoidal signal and could be from equipment at friendly units, nearby powerful electronic warfare platforms, or just a commercially available wireless device. The consequence is the disruption of existing wireless communications at the receiver side by the strong jamming signal – essentially disabling the wireless channel. Inadvertent jamming is aimless and unforeseen, but the harm caused to a wireless system, i.e. a radar system, is as severe as that from the unfriendly intentional jamming. Intentional jamming has drawn most of the attentions from the anti-jamming community [3,4] because of its importance in military applications. Therefore, research on tackling inadvertent jamming is not as well developed. In this article, we are focusing on inadvertent jamming. Traditionally, inadvertent jamming can be prevented by good spectral planning; however, due to the dramatically increasing usage of mobile wireless devices and the overcrowding of the radio spectrum, inadvertent jamming is becoming increasingly common and cannot be avoided through spectral planning alone. Since intentional jamming is usually random and noise-like, while inadvertent jamming carries real data, countermeasures developed for intentional jamming [5–7] are not effective towards inadvertent jamming. Furthermore, signal jamming is a physical layer issue that requires a solution at the physical layer. Intensive researches have been focusing on jamming detection and mitigation  in both the physical layer and Mac layer. One common anti-jamming technique in wireless communications is the spread-spectrum technique, in which a signal is transmitted on a bandwidth that is much larger than that of the frequency content of the original information [9–12]. Therefore, jamming at a specific frequency will only affect the signal during that short interval. However, the spread-spectrum technique does not involve jamming detection [13,14]. Furthermore, a large amount of bandwidth is being actively occupied even when no jamming is occurring – which is a waste of the scarce spectrum. Another technique for anti-jamming is frequency hopping [15–20], where the current communicating channel is changed after a certain duration of time no matter if there is jamming or not. However, frequency hopping has a predefined and restricted number of orthogonal channels , making it an ineffective technique for anti-jamming. A hybrid model for defending base station jamming has also been proposed  which utilize three techniques: base station replication – the unjammed base stations can serve the network, evasion of base station – no more than one base stations reach the same location at the same time, and multipath routing – at least one non-jammed path between them. This hybrid model requires additional resources, i.e. base stations, temporal slots, and multipaths, for the mitigation of jamming at the base station.
Turning to nature for a solution, weakly electric, deep ocean fish of the Eigenmannia genus generate and use electric fields for specialized active sensing that enable navigation, communication, and prey capture in the dark. Unlike human beings, Eigenmannia do not have the Federal Communications Commission to regulate spectrum allocation and use. The fish, however, have their own way of dealing with inadvertent jamming. Neuroscientists have been studying this intensively and have discovered that Eigenmannia utilize a unique neural algorithm called Jamming Avoidance Response (JAR), by which fish regulate their own emitting frequencies to avoid jamming by shifting their emitting frequency from the jamming frequency. Neuroscientists also discovered that the principle behind JAR in Eigenmannia [23–27] is based on the evaluation of the relationship between the instantaneous amplitude and phase among the local transmitting signal, the jamming signal, and the resultant beat signal between them. Due to the similarity of jamming in Eigenmannia and inadvertent jamming in our wireless systems, JAR will be an effective algorithm for combating inadvertent jamming in wireless systems like radar. Although JAR has been modeled using analog VLSI , it only works at a few hundreds of Hz frequency range. By mimicking the JAR neural algorithm with photonics [29–32], the wideband and instantaneous response typical of photonic systems, and the efficiency and unique design of the neural algorithm allow for the JAR system to respond promptly even without a priori knowledge of a jamming frequency. The proposed photonic JAR can be considered as one type of frequency hopping technique, instead of hopping at a specific time slot to a specific channel, the proposed JAR shifts away from the jammer only when the jammer is spectrally close by. Therefore, the frequency that the transmitter can escape to are virtually unlimited, with no predefined and restricted number of channels as in traditional frequency hopping techniques. In this article, we report the design and experimental demonstration of this neural-inspired photonic circuit to enable uninterrupted communications in the presence of jamming signals at a nearby frequency.
According to neuroscientists’ studies [23–27], the principle of JAR in Eigenmannia is based on the phasor phenomenon, which can be explained using Fig. 1. Suppose the fish is emitting an electric field with frequency fR, referred to as the reference signal, while another nearbyfish is emitting an electric field at fJ, acting as the jamming signal. The first electric fish receives the jamming signal alongside its own signal, generating a beat signal fB between fR and fJ. Figure 1 illustrates the relationship between fR (dashed blue) and the beat signal (solid red), when (a) fR is at a higher frequency than fI, and (b) fR is at a lower frequency than fJ.
The solid green curve represents one cycle of the beat signal envelope. The positive zero-crossing points of the beat signal (red crosses) travel around the positive zero-crossing points of the reference signal (hollow blue circles) along the beat envelope period, implying that the phase of each oscillation in the beat signal changes over the envelope period. In Fig. 1(a) where fR > fJ, the phase of the beat signal is lagging that of the reference signal at the falling edge of the beat signal envelope; while it is leading the phase of the reference signal at the rising edge of the envelope. On the other hand, in Fig. 1(b) where fR < fJ, the opposite relationship is observed - at the falling edge of the envelope, the phase of the beat signal is leading that of the reference signal; while it is lagging the phase of the reference signal at the rising edge of the envelope. Therefore, by evaluating the relationship between the instantaneous amplitude and phase of the reference signal, the jamming signal, and the resultant beat signal between them, the JAR algorithm will be able to tell whether the electric fish should tune its emitting frequency to a higher or lower frequency to avoid jamming.
3. The design of photonic JAR
In the Eigenmannia neural network, there are mainly four groups of neurons playing an important role in the JAR [23–27]. Here, we refer to them as the (1) Zero-crossing point detection unit (ZeroX unit), the (2) Phase detection unit, the (3) Amplitude unit, and the (4) Logic unit. Each group of neurons fires at different times to represent the results from each unit. Using spikes for representation has unique advantages in a biological neuron – low noise and high precision, which are essential in biological neurons due to the large number of neurons involved  in the fish for various other neural behaviors. On the other hand, mimicking a specific biological neural algorithm using photonics does not necessarily need to involve optical spikes if the function of the group of neurons can be mimicked directly by using fundamental optical phenomena in a few optical devices, eliminating the need for a large number of neurons and providing an efficient way to mimic the algorithm. To verify the photonic design, parts of the circuit were first simulated to verify design feasibility .
Figure 2 is the design of the photonic JAR, where the reference signal is a local source from a fish and the jamming signal is an external source from another. First, the referencesignal is launched to the Zero-crossing point detection unit for the identification of positive zero-crossing points in the reference signal, which will later on be used as a phase reference for the beat signal. The unit generates a spike at each positive zero-crossing point of the input signal, as illustrated by the red pulses in Fig. 2(i). Then, a portion of the reference signal, SR = sin(2πfRt), and the jamming signal, SJ = sin(2πfJt), are combined using an RF combiner where a beat signal, SB = sin(2πfRt) + sin(2πfJt), results. The beat signal and the zero-crossing spikes are launched to the Phase detection unit to determine when the beat signal is phase leading or lagging the reference signal. The phase information is represented by the modulated amplitude of the zero-crossing spikes at the Phase detection unit output. Next, amplitude rising/falling information of the beat signal is extracted by launching the beat signal to the Amplitude unit, as shown in Fig. 2(iii)-2(v). Finally, the phase and amplitude information are fed to the logic unit in which the JAR system will determine how, if necessary, to control the reference signal.
4. Photonic implementation of JAR
4.1 Zero-crossing point detection unit
The photonic implementation of JAR is mainly based on the use of semiconductor optical amplifiers (SOA), starting with the first unit – the Zero-crossing point detection unit (ZeroX unit). The goal of the ZeroX unit is to identify the positive zero-crossing points of the reference signal, which will later on be used as the phase reference in the Phase detection unit. As shown in Fig. 3, the sinusoidal reference signal from a voltage controlled oscillator (VCO) is launched to a clock divider operating at a divide-by-1 configuration before being amplitude modulated onto an optical carrier at 1549.35 nm from a distributed feedback laser. After launching to the clock divider, the reference signal becomes square-like, which enhances the detection later at the SOA.
The optical reference signal is amplified by an erbium doped fiber amplifier and is launched to the ZeroX unit, which consists of an SOA and an optical bandpass filter (OBPF 1). Self-phase modulation occurs in the SOA at each rising and falling edge of the square-like reference signal, and a ~0.4 nm broadening at the longer wavelength side (the rising edge side) is observed. An optical bandpass filter at 1549.50 nm is used to select the desired red shifted portion [34, 35], where the resultant peaks align exactly at the positive zero crossing points of the reference signal. Figure 4(a) shows the sinusoidal reference signal at various frequencies fR (top blue) and the resultant positive zero crossing pulses (bottom red) generated from the ZeroX unit. The ZeroX unit works well for reference signals from hundreds of MHz to 20 GHz due to the fast 25-ps recovery time of the SOA.
4.2 Phase detection unit
The goal of the Phase detection unit (Phase unit) is to determine whether the instantaneous phase of the beat signal is leading or lagging the phase of the reference signal . The Phase unit takes as inputs the zero-crossing reference, pulsed output from the ZeroX unit and the beat signal between the reference signal and jamming signal. The beat signal is amplitude modulated onto an optical carrier from a distributed feedback laser at 1552.57 nm. The Phase unit has the same structure as the ZeroX unit, which consists of an SOA and an optical bandpass filter (OBPF 2), but a different optical phenomenon, cross-gain modulation (XGM) , is utilized. OBPF 2 is used to extract the zero-crossing pulses at the output of the SOA after experiencing cross-gain modulation. The beat signal acts as the pump to introduce gain change in the SOA through gain depletion, and the zero-crossing reference pulses act as the probe to experience the gain change, if there is any. Since the beat signal is oscillating like a sinusoidal wave with modulated amplitude, each oscillating cycle consists of positive periods and negative periods. The optical power of the beat signal is adjusted such that only the positive period is strong enough to introduce significant gain depletion, while the “negative” period will be too weak for gain depletion.
Since the SOA has an exponential gain recovery curve, the strength of cross-gain modulation strongly depends on the temporal spacing and sequence between the pump and probe signals [37, 38]. If the instantaneous phase of the beat signal is leading the reference signal phase, the positive zero-crossing point of the reference signal will pass through the SOA after the arrival of the positive period. Thus, zero-crossing reference pulses experience significant cross-gain modulation, diminishing their amplitudes. On the other hand, if the instantaneous phase of the beat signal is lagging the reference signal phase, the positive zero-crossing point of the reference signal will align with the “negative” period of the beat signal and pass through the SOA before the arrival of the positive period. Thus, the zero-crossing reference pulses do not experience the same level of SOA gain depletion and increase in amplitude. The SOA is driven such that the recovery time is on the order of half of the reference signal period but much shorter than the beat signal frequency |fR - fJ|. Thus, independent cross-phase modulation occurs during each beat signal oscillation but not across multiple oscillation cycles.
Figure 4(b) shows the waveforms of the input beat signal (top pink) and the output zero-crossing reference pulses (bottom blue) after cross-gain modulation in the SOA. An envelope detector is used to extract the envelope of the modulated zero-crossing reference pulses (middle green). The reference signal is at 1 GHz, while the jamming signals are either 10 MHz, 50 MHz, and 100 MHz above [Fig. 4(b)i-iii] or below [Fig. 4(b)i-iii] the reference signal frequency. Pulses with amplitudes above the orange dashed line are regarded as high amplitude (considered as a binary “1”) – representing that the beat signal is phase lagging, while pulses with amplitude below the orange dashed line are regarded as low amplitude (considered as a “0”) – representing that the beat signal is phase leading. As describe in Fig. 1, the instantaneous phase of the beat signal changes between leading to lagging the reference signal phase, and a variation in the resultant reference pulses’ amplitudes are consequently expected. Therefore, when examining Fig. 4(b) where the fJ > fR, a leading in phase response (“1”) is always observed at the beat envelope’s falling edge, while a lagging in phase response (“0”) is always observed at the envelope’s rising edge. When fJ < fR an opposite relationship is observed - a lagging in phase response (“0”) is always observed at the falling edge, while a leading in phase response (“1”) is always observed at the rising edge. The above observation matches precisely with the prediction illustrated in Fig. 1, showing the accurate identification of the phase relation between the beat and reference signals.
4.3 Amplitude unit
Although the phase and amplitude relationship are observable with an oscilloscope, a unit capable of identifying the rising and falling edge of the beat envelop is necessary to ensure an autonomous JAR response, which is the major advantage of JAR over other manual anti-jamming schemes. Thus, the goal of the Amplitude unit is to return a different value for the rising and falling edge in amplitude portions of the beat envelope, i.e. a positive value output to indicate rising amplitude and a negative value output to indicate falling amplitude. The principle of the Amplitude unit is based on temporal offset and signal subtraction [39, 40], illustrated by Fig. 2(iv). Signal subtraction can be considered as signal addition with its complement, i.e. A – B = A + (-B). Therefore, when a small temporal delay is introduced in the negative version of the signal [orange curve in Fig. 2(iv)], the resultant signal will not be zero. Instead, a rising amplitude will result in a positive value while a falling amplitude will result in a negative value.
To implement rising/falling amplitude detection in the Amplitude unit, the beat signal envelope is first extracted by an envelope detector, which can be done using RF electronics due to the beat’s low frequency |fR - fJ|. The beat signal envelope is amplitude modulated onto an optical carrier from a distributed feedback laser at 1553.33 nm. Since an optical signal cannot be negative, an inverted copy of the beat signal envelope is used instead – which will be the same except with a positive offset value. Cross-gain modulation between a continuous wave light and the beat signal envelope is used to obtain the inverted version of the envelope at the continuous wavelength. Two optical bandpass filters at the corresponding wavelengths are used to separate the optical beat signal envelope and the inverted optical beat signal envelope. Then, temporal delay of about 1/12 of the beat signal period (i.e. 8.3 ns for a 10 MHz beat signal) is introduced to the inverted optical beat signal envelope before the two signals are equalized in amplitude and combined at the optical coupler. Figure 4(c) shows the combined output after a photodetector, with the beat signal envelope in brown (top) and the amplitude rising/falling detected output in blue (bottom). As shown, the rising amplitude has successfully resulted in a positive value (referred to as a “1”), while the falling amplitude has resulted in a negative value (referred to as a “0”).
4.4 Logic unit
In the Phase unit and Amplitude unit, the phase and amplitude information have been identified and encoded with a “1” or “0” correspondingly, as summarized in Table 1. The rest of the JAR system serves the purpose of (i) determining the direction of frequency tuning based on the results from previous units, and (ii) enabling or disabling the JAR depending on how close the jamming frequency is to the emitted frequency.
Examining Table 1, the relationship between amplitude, phase, and the required frequency change direction can be depicted as simple XOR logic. XOR logic can be implemented using either electronics or photonics schemes [41–44]; however, due to the low frequency nature of the Phase unit and Amplitude unit outputs (i.e. equals to |fR - fJ|, mostly below 200 MHz), electronic approaches are sufficient to implement the logic unit. The enabling of JAR is governed by whether a jamming signal is present and if the jamming signal is within the jamming frequency range, fJAR. Once the frequency adjustment is in process, the logic unit also has to determine when to stop - once the jamming signal is no longer within the jamming frequency range of the new reference signal frequency.
In this experiment, an Arduino Due board is used as the logic unit, as illustrated in Fig. 3. Basically, the logic unit supplies the voltage (VVCO) for driving the VCO that generates the desired reference frequency. Initially, the logic unit is set to output VVCO such that the VCO is generating the initial reference frequency. The JAR is enabled/disabled through the VENABLE input, which receives the output from the low pass filter following the beat signal envelope detector. Bandwidth of the low pass filter is picked to match the desired jamming frequency range fJAR; the actual fJAR can be slightly adjusted with the use of an attenuator. If |fR - fJ| < fJAR, then the beat signal envelope passes through and it serves as the enable input to VENABLE. When VENABLE is enabled, the logic unit takes the values at the VPHASE and VAMPLITUDE inputs that are connected to the outputs of the Phase unit and Amplitude unit, respectively. The logic unit performs an XOR logic operation on the VPHASE and VAMPLITUDE inputs, and returns the desired frequency shift direction, i.e. “1” represents an increase in frequency, while “0” represents a decrease in frequency. The frequency incremental step is set to be 1 MHz, which is controlled by the VVCO at the Arduino. Therefore, the VVCO will keep increasing/decreasing at a step of 1 MHz, updating the VCO driving voltage until |fR - fJ| > fJAR, at which point the VENABLE will be disabled. The finest step size that the JAR system can achieve is determined by the VCO’s tuning slope and resolution of the VCO voltage control, while the coarsest step size is determined by the tuning speed of the VCO and the spectral space available to the transmitter. The choice of step size presents a trade-off between the total time needed to escape jamming (shifting operating frequency outside of the jamming frequency range fJAR) and the spectral space that the transmitter may operate in. In a dense and heavily occupied frequency band, small step size is desired due to the difficulty for the JAR to settle on an available operating frequency using large step size. On the other hand, in a relatively sparse frequency band, a larger step size may be desired to allow faster shifting to an available frequency spot. There are two scenarios in which VENABLE will be disabled: (1) No jamming signal is present – the resultant “beat signal” will be solely the reference signal and the frequency will be too high to pass through the low pass filter; (2) If |fR - fJ| > fJAR, the jamming signal is spectrally far from the reference signal, and the beat signal envelope cannot pass through the low pass filter, which essentially disables the JAR via the VENABLE input. In either case, the logic unit will ignore the inputs at VPHASE and VAMPLITUDE, and keep the latest VVCO, i.e. no further changes are made to the VCO frequency.
5. Results - the complete photonic JAR
The jamming avoidance capability of the above proposed photonic JAR is tested using various types of jamming signals, including a pure sinusoidal wave, a sinusoidal-amplitude modulated wave with a carrier at 900 MHz and an amplitude modulation at 10 MHz, and a digitally-amplitude modulated signal with a carrier at 900 MHz and an amplitude modulation at 10 Mb/s. Figure 5(a) shows screen shots of the JAR processing when the jamming signal is a pure sinusoidal wave. As shown in Fig. 5(a) i to iii, the jamming signal first approaches from the lower frequency side (left) and moves towards the reference signal frequency. The photonic JAR is enabled once |fE - fJ| is smaller than fJAR, which pushes the reference signal to a higher frequency. Then, the jamming signal approaches from the higher frequency side (right) and moves towards the reference signal frequency, and the system pushes the reference signal to a lower frequency. Similar results are observed when a sinusoidal-amplitude modulated signal is used as the jamming signal [Fig. 5(b)] and when a digitally-amplitude modulated signal is used as the jamming signal [Fig. 5(c)]. The results show that the photonic JAR has successfully enabled automatic adjustment of the reference signal frequency to avoid jamming. Video of the photonic JAR in operation is shown in Visualization 1. It is worth noticing that the response speed of the current photonic JAR setup is mainly limited by two factors: (1) latency due to the optical path in the JAR system and (2) clock rate, processing speed, and resolution of the microprocessor (logic and decision unit). Latency could be reduced significantly to nanoseconds by integrating the optical components on a single chip, while ASIC and microprocessors with faster clock rates are commercially available to shorten the response time of the logic unit. Furthermore, moving the logic and decision tasks to the optical realm could potentially enhance the response speed of the photonic JAR.
Furthermore, a software-defined radio (SDR) is used to record the spectral waterfall – Fig. 6 on the next page shows the spectral evolution of the reference signal and jamming signal when the photonic JAR is being triggered or not, with the vertical axis representing the time evolution. As shown in Fig. 6(a)i, the reference signal fR is at 1.205 GHz while the jamming signal is approaching the reference signal from the low frequency side; however, the JAR is not triggered until fJ is higher than 1.056 GHz – which is within fJAR. The photonic JAR automatically adjusts fR to a higher frequency to keep fJ at least fJAR away from fR, and the adjustment stops once fJ stops approaching fR and is at least fJAR away. For the scenario shown in Fig. 6(a)ii where the jamming signal first approaches from the lower frequency side, stays for a short while, and then moves away – the photonic JAR will be triggered when fJ is closer than 150 MHz and will move fR away from fJ. Once the desired frequency separation is reached, fR will stay at the new frequency and will not be affected by fJ when it’s moving away from fR. The photonic JAR works similarly when the jamming signal is approaching form the higher frequency side as shown in Fig. 6(a)iii – iv. Since the instantaneous bandwidth of the SDR is only 20 MHz, the frequency tuning range is intentionally limited to within 20 MHz for displaying the whole JAR process. A similar experiment is conducted and recorded using the SDR when various modulation formats are used in the jamming signal, as shown in Fig. 6(b) and 6(c). The photonic JAR behaves similarly as in the scenario with a pure sinusoidal jamming signal, showing its effectiveness towards various types of jamming signals.
From the inspiration of the JAR neural algorithm in an electric fish, a jamming avoidance system for microwave signal is designed and experimentally demonstrated using photonics. The JAR detects whether the jamming frequency is spectrally within the jamming spectral range and intelligently moves its emitting frequency away from the jamming frequency – which will never cross the jamming signal frequency to avoid serious jamming. The photonic JAR consists of four functional units: the ZeroX unit, the Phase detection unit, the Amplitude unit, and the logic unit, which are implemented using semiconductor optical amplifiers by utilizing various optical nonlinear phenomena. The demonstrated JAR system is capable of escaping from sinusoidal, amplitude modulated, and digitally-amplitude modulated jamming signals. Due to the wideband operation capability of photonics, the photonic JAR system works well with jamming signals in the MHz range to the tens of GHz range.
National Science Foundation (NSF) (1400100 and 1653525).
The authors would like to thank Prof. Paul Prucnal from Princeton University and Dr. David Rosenbluth from Lockheed Martin for fruitful discussions at the early stage of this project.
References and links
1. P. Gupta and P. R. Kumar, “The capacity of wireless networks,” ,” IEEE Trans. Inf. Theory 46(2), 388–404 (2000). [CrossRef]
2. M. Wilhelm, I. Martinovic, J. B. Schmitt, and V. Lenders, “Short paper: reactive jamming in wireless networks: how realistic is the threat?” in Proceedings of the fourth ACM conference on Wireless network security (ACM, 2011) 47–52. [CrossRef]
3. R. A. Poisel, Modern Communications Jamming Principles and Techniques (Artech House, Inc., 2011).
4. K. Grover, A. Lim, and Q. Yang, “Jamming and anti-jamming techniques in wireless networks: a survey,” Int. J. Ad Hoc Ubiquitous Comput. 17(4), 197–215 (2014). [CrossRef]
5. Q. Xu, W. Trappe, Y. Zhang, and T. Wood, “The feasibility of launching and detecting jamming attacks in wireless networks,” in Proceedings of the 6th ACM international symposium on Mobile ad hoc networking and computing (ACM, 2005) 46–57. [CrossRef]
6. R. Gummadi, D. Wetherall, B. Greenstein, and S. Seshan, “Understanding and mitigating the impact of RF interference on 802.11 networks,” ACM SIGCOMM Comput. Commun. Rev. 37(4), 385–396 (2007). [CrossRef]
7. S. Katti, S. Gollakota, and D. Katabi, “Embracing wireless interference: analog network coding,” ACM SIGCOMM Comput. Commun. Rev. 37(4), 397–408 (2007). [CrossRef]
8. K. Grover, A. Lim, and Q. Yang, “Jamming and anti–jamming techniques in wireless networks: a survey,” Int. J. Ad Hoc Ubiquitous Comput. 17(4), 197–215 (2014). [CrossRef]
9. R. Pickholtz, D. Schilling, and L. Milstein, “Theory of spread-spectrum communications - a tutorial,” IEEE Trans. Commun. 30(5), 855–884 (1982). [CrossRef]
10. R. L. Peterson, D. E. Borth, and R. E. Ziemer, An introduction to spread-spectrum communications (Prentice-Hall, Inc., 1995) p 614.
11. C. Popper, M. Strasser, and S. Capkun, “Anti-jamming broadcast communication using uncoordinated spread spectrum techniques,” IEEE J. Sel. Areas Comm. 28(5), 703–715 (2010). [CrossRef]
12. K. M. Dostert, “Frequency-hopping spread-spectrum modulation for digital communications over electrical power lines,” IEEE J. Sel. Areas Comm. 8(4), 700–710 (1990). [CrossRef]
13. R. T. Mason, E. W. Hodges, R. C. Hemminger, and K. C. Shuey, Frequency hopping spread spectrum communications system. (Google Patents, 2005).
14. W. J. Brennan, Spread spectrum frequency hopping system and method. (Google Patents, 2002).
15. S. Khattab, D. Mosse, and R. Melhem, “Jamming mitigation in multi-radio wireless networks: Reactive or proactive?” In: Proceedings of the 4th International Conference on Security and privacy in communication networks (ACM, 2008) 27. [CrossRef]
16. S. Khattab, D. Mosse, and R. Melhem, “Modeling of the channel-hopping anti-jamming defense in multi-radio wireless networks,” In: Proceedings of the 5th Annual International Conference on Mobile and Ubiquitous Systems: Computing, Networking, and Services (Institute for Computer Sciences, Social-Informatics and Telecommunications Engineering, 2008) 25. [CrossRef]
17. V. Navda, A. Bohra, S. Ganguly, and D. Rubenstein, “Using channel hopping to increase 802.11 resilience to jamming attacks,” In: IEEE 26th IEEE International Conference on Computer Communications (IEEE, 2007) 2526–2530. [CrossRef]
18. B. Kerkez, T. Watteyne, M. Magliocco, S. Glaser, and K. Pister, “Feasibility analysis of controller design for adaptive channel hopping,” In: Proceedings of the Fourth International ICST Conference on Performance Evaluation Methodologies and Tools (Institute for Computer Sciences, Social-Informatics and Telecommunications Engineering, 2009) 76. [CrossRef]
19. H. Wang, L. Zhang, T. Li, and J. Tugnait, “Spectrally efficient jamming mitigation based on code-controlled frequency hopping,” IEEE Trans. Wirel. Commun. 10(3), 728–732 (2011). [CrossRef]
20. S. U. Yoon, R. Murawski, E. Ekici, S. Park, and Z. Mir, “Adaptive channel hopping for interference robust wireless sensor networks,” In: 2010 IEEE International Conference on Communications (IEEE, 2010) p 1–5. [CrossRef]
21. K. Pelechrinis, C. Koufogiannakis, and S. V. Krishnamurthy, “Gaming the jammer: is frequency hopping effective?” In: Proceedings of the 7th International Conference on Modeling and Optimization in Mobile, Ad Hoc, and Wireless Networks, (IEEE, 2009) 187–196. [CrossRef]
22. S. K. Jain and K. Garg, “A hybrid model of defense techniques against base station jamming attack in wireless sensor networks,” In: Proceedings of the 2009 First International Conference on Computational Intelligence, Communication Systems and Networks, (IEEE, 2009) 102–107. [CrossRef]
23. T. H. Bullock, R. H. Hamstra, and H. Scheich, “The Jamming Avoidance Response of High Frequency Electric Fish,” How do Brains Work? (Birkhäuser Boston, 1993) 509–534.
24. H. Scheich, “Neural basis of communication in the high frequency electric fish, Eigenmannia virescens (Jamming Avoidance Response),” J. Comp. Physiol. 113(2), 181–206 (1977). [CrossRef]
25. W. Heiligenberg and G. Rose, “Phase and amplitude computations in the midbrain of an electric fish: intracellular studies of neurons participating in the jamming avoidance response of Eigenmannia,” J. Neurosci. 5(2), 515–531 (1985). [CrossRef] [PubMed]
27. S. A. Stamper, M. S. Madhav, N. J. Cowan, and E. S. Fortune, “Beyond the jamming avoidance response: weakly electric fish respond to the envelope of social electrosensory signals,” J. Exp. Biol. 215(23), 4196–4207 (2012). [CrossRef] [PubMed]
28. J. E. LeMoncheck and E. John, “An analog VLSI model of the jamming avoidance response in electric fish,” IEEE J. Solid-State Circuits 27(6), 874–882 (1992). [CrossRef]
29. R. Toole and M. P. Fok, “A photonic RF jamming avoidance response system bio-inspired by Eigenmannia,” 2016 Optical Fiber Communications Conference and Exhibition (Optical Society of America, 2016), paper W1G.4. [CrossRef]
30. P. R. Prucnal, M. P. Fok, D. Rosenbluth, and K. Kravtsov, “Lightwave neuromorphic signal processing,” 2011 ICO International Conference on Information Photonics (IEEE, 2011) p. 1–2.
31. R. Toole and M. P. Fok, “Photonic implementation of a neuronal algorithm applicable towards angle of arrival detection and localization,” Opt. Express 23(12), 16133–16141 (2015). [CrossRef] [PubMed]
32. M. P. Fok, J. Ge, R. Toole, R. Lin, Q. Zhou, A. James, and A. Mathews, “Wideband dynamic microwave photonic systems: from photonics to neuromorphic,” Asia Communications and Photonics Conference 2016, (Optical Society of America, 2016), paper ATh3H.1. [CrossRef]
33. P. R. Prucnal, B. J. Shastri, T. Ferreira de Lima, M. A. Nahmias, and A. N. Tait, “Recent progress in semiconductor excitable lasers for photonic spike processing,” Adv. Opt. Photonics 8(2), 228–299 (2016). [CrossRef]
34. H. J. Lee, H. G. Kim, J. Y. Choi, and H. K. Lee, “All-optical clock recovery from NRZ data with simple NRZ-to-PRZ converter based on self-phase modulation of semiconductor optical amplifier,” Electron. Lett. 35(12), 989–990 (1999). [CrossRef]
35. L. Yin, Y. Yan, Y. Zhou, J. Wu, and J. Lin, “All-optical clock recovery from 10-Gb/s NRZ data and NRZ to RZ format conversion,” Chin. Opt. Lett. 4(1), 4–7 (2006).
36. R. Lin, L. Perea, T. Do, J. Ge, L. Xu, and M. P. Fok, “Bio-inspired optical microwave phase lock loop based on nonlinear effects in semiconductor optical amplifier,” Optical Fiber Communication Conference, (Optical Society of America, 2017), paper Th2A.41. [CrossRef]
37. Y. K. Seo, C. S. Choi, and W. Y. Choi, “All-optical signal up-conversion for radio-on-fiber applications using cross-gain modulation in semiconductor optical amplifiers,” IEEE Photonics Technol. Lett. 14(10), 1448–1450 (2002). [CrossRef]
38. R. Inohara, K. Nishimura, M. Tsurusawa, and M. Usami, “Experimental analysis of cross-phase modulation and cross-gain modulation in SOA-injecting CW assist light,” IEEE Photonics Technol. Lett. 15(9), 1192–1194 (2003). [CrossRef]
39. W. Liu, M. Li, R. S. Guzzon, E. J. Norberg, J. S. Parker, M. Lu, L. A. Coldren, and J. Yao, “A fully reconfigurable photonic integrated signal processor,” Nat. Photonics 10(3), 190–195 (2016). [CrossRef]
40. J. Xu, X. Zhang, J. Dong, D. Liu, and D. Huang, “High-speed all-optical differentiator based on a semiconductor optical amplifier and an optical filter,” Opt. Lett. 32(13), 1872–1874 (2007). [CrossRef] [PubMed]
41. J. M. Wang, S. C. Fang, and W. S. Feng, “New efficient designs for XOR and XNOR functions on the transistor level,” IEEE J. Solid-State Circuits 29(7), 780–786 (1994). [CrossRef]
42. R. P. Webb, R. J. Manning, G. D. Maxwell, and A. J. Poustie, “40 Gbit/s all-optical XOR gate based on hybrid-integrated Mach-Zehnder interferometer,” Electron. Lett. 39(1), 79–81 (2003). [CrossRef]
43. J. H. Kim, Y. M. Jhon, Y. T. Byun, S. Lee, D. H. Woo, and S. H. Kim, “All-optical XOR gate using semiconductor optical amplifiers without additional input beam,” IEEE Photonics Technol. Lett. 14(10), 1436–1438 (2002). [CrossRef]
44. M. P. Fok and P. R. Prucnal, “All-optical XOR gate with optical feedback using highly Ge-doped nonlinear fiber and a terahertz optical asymmetric demultiplexer,” Appl. Opt. 50(2), 237–241 (2011). [CrossRef] [PubMed]