Analysis and Experimental Verification of Digital Self-Interference Cancelation for Co-time Co-frequency Full-Duplex LTE

Transcription

1 nternational Journal of Signal Processing, mage Processing and Pattern Recognition, pp Analysis and Experimental Verification of Digital Self-nterference Cancelation for Co-time Co-frequency Full-Duplex LTE Qiang Xu, Xin Quan, Zhiliang Zhang, Youxi Tang and Ying Shen National Key Lab of Science and Technology on Communications, niversity of Electronic Science and Technology of China, Chengdu, China Abstract n the communication mechanism of co-time co-frequency full-duplex (CCFD), digital selfinterference cancellation (SC) is used for suppressing residual self-interference (S) after antenna and radio frequency (RF) SCs, as well as low-power multipath S. n this study, the CCFD LTE verification experiment is presented, which adopts digital SC with the S signal reconstructed in frequency domain. For the multipath Rayleigh S channel, the expression of digital SC capability is derived and the relationship among the channel estimation error, the received S power, and digital SC capability is analyzed. Experimental results show that the digital SC ability is 31. db for a 0 Hz 16QA modulated LTE S signal with frequency centered at.6 GHz and the interference-to-noise ratio of 40 db. Keywords: Co-time Co-frequency Full-Duplex; Digital Self-interference; LTE 1. ntroduction To improve the spectral efficiency of frequency division duplex or time division duplex system, theoretical analysis and preliminary experimental verification for co-time cofrequency full-duplex (CCFD) were carried out by industrial and academic scientists in recent years [1-3]. Researches have shown that CCFD allows the transmitter and the receiver of an equipment work simultaneously using the same frequency. Therefore, uplink and downlink of wireless communication can use the same frequency resource simultaneously. n theory, CCFD can double the spectral efficiency [4]. As the transmitter and the receiver work simultaneously at the same frequency, the transmitted signal of CCFD transmitter will cause strong interference to the local receiver. Therefore, the critical step of CCFD design is to suppress the strong self-interference (S) [1, 3]. Studies on self-interference cancellation (SC) carried out in recent years can be categorized as antenna SC [5, 6], radio frequency (RF) SC [7-9], and digital SC [3, 10-1]. Among them, the antenna SC and the RF SC are implemented in the analog domain. They are used to eliminate the line of sight (LOS) S [7] and some strong multipath S [13], ensuring the received signal could pass the analog to digital converter (ADC). The performance of the cancellation is related to S signal bandwidth, adjustment error, RF channel non-ideal characteristics, and equipment environment change [14, 15]. After RF SC, digital SC is performed in the digital domain, as an important composition part of SC. t can significantly suppress the residual multipath S, some low-power multipath S, and the residual S caused by the RF adjustment error and environmental changes [15]. Digital SC scheme can be accomplished by employing S signal reconstructing [10, 1, 16], adaptive filtering [3, 17] and pre-coding [4, 18], and has some preliminary engineering Book made by this file is LLEGAL. SSN: JSP Copyright c 014 SERSC

2 nternational Journal of Signal Processing, mage Processing and Pattern Recognition verification [1, 4, 7]. Among them, a typical S reconstructing digital SC scheme is given in [7] and [16]. n [7], RF SC and digital SC were used sequentially to reduce the S power. Digital SC was carried out in three steps, i.e., (1)using the least squares (LS) method to perform channel estimation, ()reconstructing the S signal in time domain by employing channel estimation and the digital transmitted signal, (3)subtracting the reconstructed S signal from the received signal. For a 10 Hz WiFi S signal with the received power varying from 30 dbm to 10 dbm, the digital SC capability was about 30 db. The experiment verified the digital SC capability at high interference-to-noise ratio (NR), however it did not reflect the relationship between digital SC capability and received S power. And the effect of channel estimation error on the digital SC capability was not considered. n [16], SC scheme was organized by, (1)generating the digital cancelling signal by employing the channel estimation and digital transmitted signal in the frequency domain, ()transmitting the digital cancelling signal through the RF path to obtain the RF cancelling signal, (3)subtracting the RF cancelling signal and digital cancelling signal sequentially from the received signal to accomplished RF and digital SCs. The experiments examined the RF SC capability and the joint cancellation capability under different received S power. For a 10 Hz WiFi S signal, the maximum joint cancellation was 36 db. Due to the limitation of the SC scheme, the digital SC capability changed with the RF SC capability. Besides, the experiment lacked the independent verification of the digital SC capability. Therefore, the relationship among the received S power, the channel estimation error, and the digital SC capability needs to be explored and experimentally verified. n this study, the CCFD LTE verification experiment is presented, which adopts digital SC with the S signal reconstructed in frequency domain after applying RF SC proposed in [7]. nder the condition of multipath S Rayleigh channel, the systematic scheme and the cancellation capability of the digital SC used in experiment are analyzed. A closed-form expression for digital SC capability is derived in terms of the received S power and the S channel estimate error. Simulation and experimental curves are presented to show that the digital SC capability decreases with the increase of channel estimation error, and increases with the rising of received S power. The trends and differences among these curves are also analyzed. The rest of this paper is arranged as follows. Section gives the systematic scheme of digital SC in our experiment, and then presents the error analysis and theoretical and simulation results. Section 3 introduces the platform, procedure, and key algorithm used in the experiment. Section 4 gives experimental results and analyzes the difference between experimental and simulation results. Section 5 is the summary.. Principle of Experiment Design.1. Systematic Scheme Book made by this file is LLEGAL. An OFD-based CCFD single-input single-output (SSO) system is shown in Figure 1, where denotes the carrier frequency. Take the local equipment as an example. Since the transmitter and the receiver work at the same time and the same frequency, the receiver hears not only the desired signal r () t from the remote equipment, but also the S signal r () t from its transmitter. Therefore, it is essential to cancel S signal. To improve the SC performance, we employ both RF SC and digital SC to reduce S power. RF SC aims to eliminating highpower LOS S, and digital SC eliminates the residual multipath S. n this paper, we focus on the digital SC, and the RF SC can be found in [14]. According to the analysis above, the signal rt () received by the local receiver is: 300 Copyright c 014 SERS

3 nternational Journal of Signal Processing, mage Processing and Pattern Recognition r ( t ) r ( t ) r ( t ) n ( t ) (1) where r () t, r () t, nt () are the desired signal, the S signal, and the white Gaussian noise, respectively. As there is a LOS path between the transmit antenna and the receive antenna, the S channel response is modeled as Rice channel, thus the S signal r () t can be expressed as: L r ( t) a ( t) s ( t t ) a ( t) s ( t t ) 0 0 () l l where s () t is the local transmitted signal and L is the multipath number of the S channel. a 0 () t and t 0 are the attenuation factor and delay of LOS path at time t, respectively. a () t and t are the attenuation factor and delay of the l path at time t, respectively. l Baseband Tx Data Baseband Rx Data QA odulation B ( k) Digital SC S H( k) S Channel Regeneration Estimation Yk ( ) S ( C k ) + Rk ( ) FFT OFD odulation CP Removal Local Equipment DAC ADC l 1 Tx Radio Rx Radio RF SC TX RX f 0 r () t rt () ht () f 0 r () t f 0 RX TX Remote Equipment Figure 1. Systematic Scheme of Digital SC in SSO CCFD System Because most of the LOS S power can be reduced by RF SC, the S channel impulse response after RF SC can be modeled as Rayleigh channel, and thus the residual S after RF SC is: L r ( t) a ( ) ( t ) D t s t (3) l l l 1 After applying sequentially the analog-to-digital, cyclic prefix (CP) removal, and fast Fourier transform (FFT), the digital frequency domain received signal is obtained, where the symbol at the k-th subcarrier is R ( k ) R ( k ) R ( k ) N ( k ) (4) Book made by this file is LLEGAL. with R ( k ), R ( k ), and Nk ( ) the digital frequency domain symbols of r () t, r () t, and nt (), respectively. Supposing the subcarrier number of the system is and each subcarrier has the equal power, after ideal synchronization, R ( k ) and R ( k ) can be expressed as E S R ( k) H ( k) B ( k) (5) D l Copyright c 014 SERSC 301

4 nternational Journal of Signal Processing, mage Processing and Pattern Recognition E S R ( k) H ( k) B ( k) (6) where E and E are the transmit power and desired signal power, respectively. H ( k ) S S and H ( k ) are the frequency domain channel response of the k-th subcarrier of the S channel and the desired signal channel, respectively. B ( k ) and B ( k ) are the frequency domain transmitted signal of the kth subcarrier of the S signal and the desired signal, respectively. Digital SC is accomplished by employing the frequency domain reconstructed S signal [7]. As shown in Figure 1, it consists of three components. First, apply S channel estimation to obtain the frequency domain channel response Hk ) ( ). Then, employ the channel estimation ) Hk ( ) and the frequency domain transmitted signal B ( k ) to obtain the frequency domain reconstructed self-interference signal S ( k ), C E ) S S ( k) H( k) B ( k) (7) C Finally, subtract the reconstructed S signal S ( k ) from Rk ( ) in the frequency domain to obtain the digital signal Y ( k) after digital SC. Thus Y ( k ) is expressed as r C Y ( k) R( k) S ( k) R ( k) R ( k) N ( k) (8) C r where R ( k) is the residual S symbol after digital SC: E ) S R ( k) R ( k) S ( k) H( k) H( k) B ( k) (9) r C The performance of the digital SC can be measured by the digital SC capability G (db) [14], which is defined as: E s G 10 lg E s where E is the power of the S symbol Rk ( ) before digital SC, and residual S symbols R ( k ) after digital SC... Error Analysis r r (10) E is the power of r Book made by this file is LLEGAL. From Eq. (9), it can be seen that after ideal synchronization, the residual S signal is determined by S channel estimation Hk ˆ ( ), thus the accuracy of Hk ˆ ( ) is the key factor influencing the digital SC capability. According to Eq. (9), the power of residual S E is: r 30 Copyright c 014 SERS

5 nternational Journal of Signal Processing, mage Processing and Pattern Recognition E ) S E E ( ) ( ) ( ) r H k H k B k E ) S E H ( k ) H ( k ) ge S where g is the mean square error (SE) of channel estimation defined as t can be seen that channel estimation g. r ) (11) g E H( k) H( k) (1) E is a function of the transmitted S power Substituting Eq. (11) into Eq. (10), we have the digital SC capability : E s R N G 10 lg 10 lg ge gr S s N 1 1 E and the SE of S where R is the interference-to-noise ratio (NR), i.e., the power ratio of the S signal before N digital SC and the thermal noise, and can be expressed as N s (13) E R (14) g is the normalized mean square error (NSE) of channel estimation, defined as g g (15) E H ( k) We also analyze the influence of residual S on the desired signal demodulation. After digital SC, the desired signal-to-interference and noise ratio is R, SN where R S R SN E R N ge g R R R s Book made by this file is LLEGAL. S N S S is the power ratio of the S signal and the desired signal before digital SC: R E (16) (17) S E According to the bit error rate (BER) function of 16QA under Rayleigh channel [19], the demodulation BER of the desired signal is: Copyright c 014 SERSC 303

6 nternational Journal of Signal Processing, mage Processing and Pattern Recognition P e 1 R 9 R 4.5 R SN SN SN arct an 4.5 R 16.5 R p R SN SN SN (18).3. Theoretical and Simulation Results On the basis of Sections.1 and., this section presents theoretical analysis and simulation verification to measure the influence of the channel estimation error and the received S power on digital SC capability and BER of the desired signal. Parameters used in analysis and simulations are shown in Table 1. Theoretical analysis parameters Simulation parameters Table 1. Analysis and Simulation Parameters odulation mode Desired signal channel S channel Desired signal power Thermal noise power Received S power (dbm) S channel estimation NSE 16QA Rayleigh channel Rayleigh channel 68 dbm 98 dbm OFD subcarrier number 048 odulation mode 88, 78, 68, 58, 48 0, 10 4, 10 3, 10, QA FFT length 048 Signal bandwidth Desired signal channel S channel SNR 0 Hz Rayleigh channel Rayleigh channel 30 db NR (db) 10, 0, 30, 40, 50 S channel estimation NSE Desired signal channel estimation 0, 10 4, 10 3, 10, 10 1 deal estimation According to Eq. (13), the relationship between the digital SC capability and the received S power can be obtained different S channel estimation NSE. Theoretical and simulation results are shown in Figure with the S channel estimation NSE ranging from 0 to The simulation curve matches well with the theoretical curve. t can be seen that: Book made by this file is LLEGAL. (1) For a specific NSE, the digital SC capability G increases with the increase of the received S power and finally reaches an upper bound. The upper bound can be explained by obtain the limit of as R approaches infinity. N G lim R lim 10lg gr NR N N 1 10lg g 1 (19) 304 Copyright c 014 SERS

7 nternational Journal of Signal Processing, mage Processing and Pattern Recognition t can be seen that the upper bound of the digital SC capability is only related to the S channel estimation NSE. () For a given received S power, i.e., NR R is fixed, the digital SC capability G N decreases as the S channel estimation NSE increases. Digital SC capability G(dB) Simulation NSE=0 Simulation NSE=10 4 Simulation NSE=10 3 Simulation NSE=10 Simulation NSE=10 1 Analysis Results NR (db) Figure. Digital SC Capability versus the Received S Power According to Eqs. (16) and (18), the relationship between the demodulation BER of the desired signal and the received S power can be obtained. The theoretical and simulation results are shown in Figure 3 with the S channel estimation NSE ranging from 0 to t can be seen that: Digital SC capability G(dB) Simulation NSE=0 Simulation NSE=10 4 Simulation NSE=10 3 Simulation NSE=10 Simulation NSE=10 1 Analysis Results Book made by this file is LLEGAL NR (db) Figure 3. Demodulation BER of the Desired Signal versus the Received S Power Copyright c 014 SERSC 305

8 nternational Journal of Signal Processing, mage Processing and Pattern Recognition (1) For a specific S channel estimation NSE, the demodulation BER increases with the increase of the received S power, which indicates that a stronger S produces a larger impact on the desired signal demodulation. () For a given NR R N NSE increases. 3. Experimental Verification Scheme, the demodulation BER increases as the S channel estimation We built a test platform to experiment the validity of the digital SC scheme and the effect of the S channel estimation error and the received S power on digital SC Experimental Platform and Environment The experiments were accomplished indoors using the test platform we built. The experimental environment is shown in Figure 4, and some key parameters of the platform are shown in Table. (a) Panorama (b) Enlarged Details Figure 4. Experimental Environment Table. Key Parameters Parameter ndicator Antenna configuration SSO odulation mode 16QA+OFD Signal bandwidth 0 Hz Transmit intermediate frequency 1.88 Hz DA bits 16 DA sampling rate Hz RF carrier frequency.6 GHz Receive intermediate frequency Hz AD nominal number of bits 14 AD effective number of bits 1 AD sampling rate 1.88 Hz Book made by this file is LLEGAL. 306 Copyright c 014 SERS

9 nternational Journal of Signal Processing, mage Processing and Pattern Recognition 3.. Experiment Procedure Our goal is to verify the validity of digital SC and the impact of the received S power on the digital SC capability and the demodulation BER of the desired signal. To improve the SC performance, we employ both RF SC and digital SC to reduce S power. The experiments are organized as follows. (1) Record the power of noise without the S signal or the desired signal. () ake the remote equipment transmit signals and adjust the transmitting power to ensure the SNR after ADC equal 30 db. (3) Stop the remote equipment transmitting. (4) ake the local equipment transmit signals and adjust the transmitting power to ensure the NR after ADC equal 10dB. (5) Record the difference G between the power before and after digital SC. 1 (6) ake the remote equipment transmit and record the demodulation BER P. e1 (7) Adjust the local equipment transmitting power to set the NR equal 14, 0, 7, 30, 34, and 40 db, respectively, repeat (3) to (6), and record the digital SC capability G and the demodulation BER P under different NR. ei 3.3. Key Algorithms The key algorithms in experiments include the synchronization algorithm and the S channel estimation algorithm. A brief introduction is listed as follows. Local PSS Sequ ence Conjugating FFT Start Received Signal Narrow-band Filtering Down- Sampling Cross-correlation Calcu lating Book made by this file is LLEGAL. Exceed Threshold? Figure 5. The S Signal Synchronization in CCFD Experiments Synchronization: n the 3GPP LTE protocol, a primary synchronization signal (PSS) sequence, which is generated by a Zadoff Chu (ZC) sequence, has an excellent autocorrelation and cross-correlation properties. Thus, the local equipment and the remote End yes no i Copyright c 014 SERSC 307

10 nternational Journal of Signal Processing, mage Processing and Pattern Recognition equipment adopt different PSS sequences in the experiments, and cross-correlation algorithm is used to synchronization. The process of synchronization is shown in Figure 5. After the received signal is filtered and downsampled, it takes sliding correlation with the local PSS sequence. Synchronization is completed when the correlation result exceed the threshold. Otherwise, synchronization continues S Channel Estimation: We use the LS algorithm to obtain the S channel estimation by exploiting the reference signal in frequency domain. The algorithm is implemented as ) Y ( k) Hk ( ) (0) X( k) ) where Y ( k ), X( k ), and Hk ( ) are the received reference signal, the transmitted reference signal, and channel estimation in frequency domain, respectively. 4. Experimental Results and Analysis 4.1. Experimental Results The curves of the digital SC capability and the BER of the desired signal are shown in Figure 6 and Figure 7, with a SNR of 30dB and the received S power varying from 10 db to 40 db. Comparing experimental curves with simulation curves, it can be seen that: (1) Experimental curves have the same trend with simulation curves. Both digital SC capability and the BER of the desired signal increase with the increase of the received S power, and digital SC capability tends to an upper bound. () Experimental results with the NR varying from 10 db to 30 db quite match simulation results with NSE varying from 10 1 to 10 3, correspondingly. While there s a significant gap between the experimental result with the NR of 40 db and the simulation result with NSE of Digital SC capability G(dB) Experimental Result Simulation NSE=10 4 Simulation NSE=10 3 Simulation NSE=10 Simulation NSE=10 1 Book made by this file is LLEGAL NR (db) Figure 6. Digital SC Capability versus the Received S Power 308 Copyright c 014 SERS

11 nternational Journal of Signal Processing, mage Processing and Pattern Recognition seful signal demodulation BER Experimental Result Simulation NSE=10 4 Simulation NSE=10 3 Simulation NSE=10 Simulation NSE= NR (db) Figure 7. BER of the Desired Signal versus the Received S Power The S signal spectra before and after digital SC with the NR of 30 db are shown in Figure 8, where the S power is normalized to noise power. The S spectrum fluctuates before digital SC, whereas it becomes relatively even with an NR of 3.15 db after digital SC. t can be concluded that the digital SC capability is about 6.85 db, and digital SC effectively improves the unevenness in the spectrum caused by RF SC. Normalized Power (db) Before Digital SC After Digital SC Book made by this file is LLEGAL Frequency(Hz) Figure 8. Signal Spectra before and after Digital SC with NR of 30 db Copyright c 014 SERSC 309

12 nternational Journal of Signal Processing, mage Processing and Pattern Recognition 4.. Comparison and Analysis Some analysis on the difference between experimental results and simulation results are given as follows. n experiments, the S channel estimation is obtained by employing LS algorithm. sing Eq.(0), we have ) Nk ( ) H( k) H( k) (1) X( k) Substituting Eq. (1) into Eq. (15), the LS channel estimation NSE can be obtained as N ( k) X ( k) s 1 g E H( k) E R N According to Eq. (), the S channel estimation NSE is 10 1, 10, 10 3 and 10 4 with NR varying from 10 db to 40 db, respectively. Therefore, experimental results with the NR lower than 30 db quite match simulation results. While in the case of NR being 40 db, the experimental result is dramatically different with the simulation result when NSE is The major reason of the dramatic difference is that, in the case of high NR, non-ideal factors such as phase noise and amplifier nonlinearity, dominantly influence the digital SC and thus the demodulation BER of the desired signal. 5. Conclusion n this paper, we designed the CCFD LTE verification experiment by adopting the digital SC with the S signal reconstructed in frequency domain. nder the condition of multipath S Rayleigh channel, the closed-form expression for digital SC capability was derived. Simulation and experimental curves were presented to show that, the digital SC capability decreases with the increase of the channel estimation error, and increases with the rising of the received S power. The analysis and experimental results in this paper provided theoretical guidance for the algorithm selection and optimization in digital SC, and the reference data for CCFD LTE engineering application. Acknowledgements This work was supported in part by the National Science Foundation of China (under grant NO , /L05, , and ), the National Science and Technology ajor Project (under grant NO. 014ZX , 01ZX , and 011ZX ) Book made by this file is LLEGAL. References [1]. Duarte and A. Sabharwal, Full-duplex Wireless Communication sing Off-the Shelf Radios: Feasibility and First Results, Conference Record of the Forty Fourth Asilomar Conference on Signals, Systems and Computers, California, SA, (010) November [] A. Sahai, G. Patel and A. Sabharwal, Pushing the Limits of Full-Duplex: Design an Real-time mplementation, The Computing Research Repository, (010). [3] R. Lopez-Valcarce, E. Antonio-Rodriguez, C. osquera and F. Perez-Gonzalez, An Adaptive Feedback Canceller for Full-Duplex Relays Based on Spectrum Shaping, EEE Journal on Selected Areas in Communications, vol. 30, (01), pp [4] Y. Hua, P. Liang, Y. a and A. C. Cirik, A ethod for Broadband Full-Duplex O Radio, EEE Signal Processing Letters, vol. 19, (01), pp () 310 Copyright c 014 SERS

13 nternational Journal of Signal Processing, mage Processing and Pattern Recognition [5] J.. Choi,. Jain, K. Srinivasan, P. Levis and S. Katti, Achieving Single Channel, Full Duplex Wireless Communication, obicom 10 Proceedings of the 17th annual international conference on obile computing and networking, llinois, SA, (010) September 0-4. [6]. A. Khojastepour, K. Sundaresan, S. Rangarajan, X. Zhang and S. Barghi, The Case for Antenna Cancellation for Scalable Full-Duplex Wireless Communications, 10th AC Workshop on Hot Topics in Networks, assachusetts, K, (011) November [7]. Jain, J.. Choi, T. Kim, D. Bharadia, S. Seth, K. Srinivasan, P. Levis, S. Katti and P. Sinha, Practical, Real-time, Full Duplex Wireless, obicom '11 Proceedings of the 17th annual international conference on obile computing and networking, Nevada, SA, (011) September [8] S. Hong, J. ehlman and S. Katti, Picasso: Flexible RF and Spectrum Slicing, Proceedings of the AC SGCO 01 conference on Applications, technologies, architectures, and protocols for computer communication, Helsinki, Finland, (01) August [9]. E. Knox, Single Antenna Full Duplex Communications using a Common Carrier, 01 EEE 13th Annual Wireless and icrowave Technology Conference, Florida, SA, (01) April [10] Y. J. Lee, J. B. Lee, S.. Park, Y. T. Lee, H.. Kim and H. N. Kim, Feedback Cancellation for T-DB Repeaters based on Frequency-domain Channel Estimation, EEE Transactions on Broadcasting, vol. 57, (011), pp [11] Y. Liu, X. Xia and H. Zhang, Distributed Space-Time Coding for Full-Duplex Asynchronous Cooperative Communications, EEE Transactions on Wireless Communications, vol. 11, (01), pp [1] D. Chang, Apparatus and ethod for Removing Self-nterference and Relay System for the Same,.S. Patent S844B, (01) June 17. [13] J. G. cichael, K. E. Kolodziej, Optimal tuning of analog self-interference cancellers for full-duplex wireless communication, 01 50th Annual Allerton Conference on Communication, Control, and Computing, llinois, SA, (01) October 1-5. [14] Q. Xu, X. Quan, W. Pan, S. Shao and Y. Tang, Analysis and Experimental Verification of RF Self- nterference Cancelation for Co-time Co-frequency Full-Duplex LTE, Journal of Electronics & nformation Technology, to appear, (013). [15] G. R. Kenworthy, Self-cancelling full-duplex RF communication system,.s. Patent S , (1997) November 5. [16]. Duarte, C. Dick and A. Sabharwal, Experiment-driven Characterization of Full-Duplex Wireless Systems, EEE Transactions on Wireless Communications, vol. 11, (01), pp [17] C. C. Tung, Full Duplex Wireless ethod and Apparatus,.S. Patent S A1, (01) October 18. [18] D. Choi and D. Park, Effective self-interference cancellation in full duplex relay systems, Electronics Letters, vol. 48, (01), pp [19]. K. Simon and. S. Alouini, Digital communication over fading channels, John Wiley & Sons, New York, (000), pp. 1. Authors Qiang Xu was born in Sichuan, China, in He received the B.E. degree in Communication and nformation Engineering from the niversity of Electronic Science and Technology of China, Chengdu, China, in 006 and the.s. degrees in communications and information systems from the niversity of Electronic Science and Technology of China, Chengdu, China, in 009. Since 009, he has been with the National Key Laboratory of Science and Technology on Communications, niversity of Electronic Science and Technology of China, as an Assistant Lecturer. His general research interests include hardware and software design for communications that touch the physical world, and wireless mobile systems with emphasis on signal processing in communications. Book made by this file is LLEGAL. Copyright c 014 SERSC 311

14 nternational Journal of Signal Processing, mage Processing and Pattern Recognition Xu Qiang was born in Hebei, China, in She received the B.E. degree in Electronic and nformation Engineering from Yanshan niversity in 010. She is currently pursuing the Ph.D. in communication and information systems at the niversity of Electronic Science and Technology of China, Chengdu, China. Her research interests include full-duplex communication and signal processing. O m nl ad in e ev by e th rsio is n fil O e is nly L. LE G AL. Zhiliang Zhang was born in Guangdong, China, in He received the B.E. degree in electronic information science and technology and.s. degree in circuits and systems from the Sichuan niversity, Chengdu, China, in 003 and 006, respectively. Since 010, he has been working toward the Ph.D. degree in communications and information systems at the niversity of Electronic Science and Technology of China, Chengdu, China. His general research interests include full-duplex communication and signal processing. Bo ok Youxi Tang, is a professor with the National Key Laboratory of Science and Technology on Communications, niversity of Electronic Science and Technology of China. He received the B.E. degree in radar engineering from College of PLA Ordnance, in 1985 and the.s. and Ph.D. degrees in communications and information systems from the niversity of Electronic Science and Technology of China, in1993 and 1997, respectively. He His research interests include spread spectrum systems and wireless mobile systems with emphasis on signal processing in communications 31 Ying Shen received the BS, S, and PhD degrees in communications and information systems from the niversity of Electronic Science and Technology of China, Chengdu, China, in 00, 006, and 009, respectively. Currently, he is working at the National Key Laboratory of Science and Technology On Communications, niversity of Electronic Science and Technology of China, Chengdu, China. His research interests include cognitive radio, O and broadcasting systems. Copyright 014 SERS