ARTICLE IN PRESS Signal Processing 85 (2005) 2233–2303 www.elsevier.com/locate/sigpro Bibliography on cyclostationarity Erchin Serpedina,Ã, Flaviu Pandurua, Ilkay Sarıa, Georgios B. Giannakisb aDepartment of Electrical Engineering, Texas A&M University, College Station, TX 77843-3128, USA bDepartment of Electrical and Computer Engineering, University of Minnesota, 200 Union Street SE, Minneapolis, MN 55455, USA Received 19 August 2004; received in revised form 9 May 2005 Available online 2 June 2005 Abstract The present bibliography represents a comprehensive list of references on cyclostationarity and its applications. An attempt has been made to make this bibliography complete by listing most of the existing references up to the year 2005 and by providing a detailed classification group. r 2005 Elsevier B.V. All rights reserved. Keywords: Cyclostationarity; Cyclostationary; Periodically correlated processes; Cyclic correlation; Spectral correlation; Cyclic moments; Higher-order cyclostationarity Contents Introduction 2235 1. Statistical theory of cyclostationarity 2236 1.1. Theory of periodically and almost periodically correlated processes 2236 1.2. Stochastic processes theory 2236 1.3. Time series theory 2236 1.4. Ergodic theory of cyclostationary sequences 2236 1.5. Tests for cyclostationarity 2237 1.6. Stationarization 2237 1.7. EOF—empirical orthogonal functions 2237 1.8. Random fields 2237 1.9. Law of large numbers 2237 ÃCorresponding author. Tel.: +1 979 458 2287; fax: +1 979 862 4630. E-mail addresses: [email protected], [email protected] (E. Serpedin), [email protected] (F. Panduru), [email protected] (I. Sarı), [email protected] (G.B. Giannakis). 0165-1684/$ - see front matter r 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.sigpro.2005.05.002 ARTICLE IN PRESS 2234 E. Serpedin et al. / Signal Processing 85 (2005) 2233–2303 1.10. Asymptotic normality 2237 1.11. Second and higher-order statistics 2237 2. Signal processing 2237 2.1. Estimation 2237 2.2. Detection 2239 2.3. System identification 2239 2.4. Prediction 2239 2.5. Period analysis 2239 2.6. Signal processing systems 2239 2.7. Image processing 2240 3. Communications 2240 3.1. Modeling and transforms 2240 3.2. Modulation, multiple access and coding 2240 3.3. Noise and interference 2241 3.4. Channels 2241 3.5. Equalization 2241 3.6. Filtering 2242 3.7. Algorithms 2242 3.8. Cellular and microcellular systems 2242 3.9. Military communications 2242 4. Antenna array processing 2243 4.1. Adaptive arrays 2243 4.2. MIMO systems 2243 4.3. Beamforming 2243 4.4. DOA—direction of arrival 2243 4.5. TDOA—time difference of arrival 2243 5. Mechanics 2243 5.1. Mechanical vibrations 2243 5.2. Rotating mechanisms 2243 6. Oceanography and hydrology 2243 7. Climatology and meteorology 2243 8. Economics 2243 9. Astronomy and satellite communications 2244 10. Magnetism and electromagnetism 2244 11. Geography, seismology and environment 2244 12. Medicine, biology 2244 13. Optics 2244 14. Acoustics and speech 2244 15. Networks 2244 15.1. Telecommunications and computer networks 2244 15.2. Subscriber lines 2244 15.3. Power lines 2244 15.4. Queueing 2244 15.5. Neural networks 2244 15.6. ATM networks 2244 16. Electronics 2244 16.1. RF circuits 2244 16.2. Switched-capacitor networks 2244 ARTICLE IN PRESS E. Serpedin et al. / Signal Processing 85 (2005) 2233–2303 2235 16.3. PLL—phase locked loop 2244 16.4. Integrated circuits and semiconductors 2244 16.5. Cyclostationary noise in mixers and oscillators 2245 17. Books on cyclostationarity 2245 18. Theses and dissertations on cyclostationarity 2245 19. Miscellaneous 2245 References 2245 Introduction cesses. In literature, cyclostationary processes are named in multiple different ways such as periodi- Cyclostationary processes are those signals cally correlated, periodically nonstationary, peri- whose statistics vary almost periodically, and they odically nonstationary or cyclic correlated are present in numerous physical and man-made processes. processes: ultrasonic imaging of materials and Historically, it appears that Bennett (1958) [77] biological tissues, medicine (EEG, ECG, circadian observed for the first time the presence of rhythm), solid state and plasma physics, radio- cyclostationary signals in the design of synchroni- astronomy, mechanics (vibration and noise analy- zation algorithms for communications systems. sis for condition-based monitoring of rotating Shortly after (1959–1980), several mathematicians machineries: engines, turbines), radar, sonar, from the former Soviet Union (Gladyshev, Gud- telemetry, and communications systems, modeling zenko, Dragan, etc.) introduced key concepts for and performance evaluation of the noise figure in representation of cyclostationary processes electronic and optic devices, etc. In communica- [345–350,556–559,585,586]. More specifically, in tion systems, operations like sampling, modula- 1959 Gudzenko [586], presented a study on non- tion, mixing, multiplexing, coding, and scanning parametric spectral estimation of cyclostationary create an information bearing signal with periodic processes. Later in 1961 and 1963, Gladyshev or almost periodic characteristics; in ultrasonic [556,559], worked on spectral analysis recognizing imaging, regular scatterer spacings induce a quasi- relation between periodically correlated processes periodicity on ultrasonic pulse echo scans; in and stationary vector sequences, and he also electronics, noise at the output of a nonlinear introduced the concept of almost periodically electronic device is excited by periodic signals correlated processes. In 1963, Nedoma [1005] (noise currents in MOSFET vary periodically with presented cycloergodicity for cyclostationary pro- the oscillating waveform) and many circuits cesses with single period and later in 1983, Boyles present time-varying operating points (mixers, and Gardner [145] extended it to general cyclosta- oscillators, samplers, and switched filters); in tionary processes with multiple periods. After climatology, presence of rhythmic or seasonal Bennett’s first usage of cyclostationarity in com- behavior in nature results in repetitive climatolo- munication context, Franks (1969) [424] devoted a gical data; in rotating machinery vibration signals relatively detailed section of his book to cyclosta- produced by IC engines have cyclic nature. In tionarity in communication. Then, in 1969 Hurd’s short, cyclostationary signals are frequently en- thesis [663] appeared as a very good introduction countered in a broad range of applications and to continuous time cyclostationary processes. In since exploitation of the periodic features present 1975, Gardner and Franks [449] studied benefits of in cyclostationary signals generally leads to algo- series representation of cyclostationary processes rithms with substantially improved performance especially in the context of optimum filtering. First relative to the case when the processed signals are comprehensive treatment of cyclostationarity in viewed as stationary, cyclostationary signals ap- communication and signal processing appeared in pear as the most suited framework for modeling Gardner’s book [452] in 1985. In 1987, Gardner and processing such periodically correlated pro- presented his nonprobabilistic statistical theory of ARTICLE IN PRESS 2236 E. Serpedin et al. / Signal Processing 85 (2005) 2233–2303 cyclostationarity in [461]. In parallel, Giannakis Despite authors’ huge efforts to include all the and Dandawate [295,297,547], approached cyclos- existing references that deal with cyclostationarity, a tationarity within the framework of stochastic number of references might have not been included. processes. In 1992, Spooner [1272] considered the We would like to apologize in advance to all the theory of higher-order cyclostationarity. researchers whose works have not been cited in this As the theory of cyclostationary developed, lots bibliography. of related works appeared in many different areas such as climatology in Hurd [676], hydrology in Classification Kacimov [763], medicine and biology in Finelli [413], oceanology in Dragan [352], economics in 1. Statistical theory of cyclostationarity Pagano [1047], mechanics in Sherman [1241] and many fields in communication and signal proces- 1.1. Theory of periodically and almost periodically sing like crosstalk in Campbell [173] and para- correlated processes meter estimation in Gardner [461]. Also, the last decade marked a renewed interest in cyclostatio- [28] [35] [36] [42] [77] [80] [95] [134] [136] [261] narity through the pioneering works of Tong [280] [312] [314] [317] [320] [345] [347] [348] [353] [1324], [1327], Tugnait [1361], Ding [333] and [359] [373] [393] [400] [473] [491] [547] [557] [558] Giannakis [547], which generated an intensive [640] [642] [665] [670] [680] [686] [689] [712] [731] research activity in the area of blind estimation [733] [735] [766] [772] [778] [804] [805] [843] [909] and equalization of communications channels. [913] [967] [970] [1007] [1058] [1134] [1137] [1147] In short, after the early treatment, mainly two [1165] [1225] [1255] [1272] [1279] [1280] [1281] research groups have contributed significantly in [1310] [1392] [1396] [1446] [1496] [1497] [1501] USA to the theory and applications of cyclosta- [1506] tionary signal processing in the engineering com- munity, namely the research centers of Professors 1.2. Stochastic processes theory W. Gardner and G.B. Giannakis. Basically, Gardner builds the theory of CS signals within a [26] [27] [30] [36] [37] [42]