## Suppression of Multiple Narrowband Interferers in a Spread-spectrum Communication System (2000)

Venue: | IEEE J. Select. Areas Commun |

Citations: | 3 - 0 self |

### BibTeX

@ARTICLE{Carlemalm00suppressionof,

author = {C. Carlemalm and H.V. Poor and A. Logothetis},

title = {Suppression of Multiple Narrowband Interferers in a Spread-spectrum Communication System},

journal = {IEEE J. Select. Areas Commun},

year = {2000},

volume = {18},

pages = {1365--1374}

}

### OpenURL

### Abstract

We consider the problem of estimating and suppressing many unknown independent and time-varying interferers in a spread-spectrum communication system. The interferers are assumed to be present in a wide frequency range. In order to detect, estimate and track the interference, we use a bank of hidden Markov model lters operating in the frequency domain. The hidden Markov model lters' outputs are then used to suppress the existing interference. The computational complexity of our scheme is only linear in the number of interferers. The simulation studies show that our proposed novel schemes adapt quickly in tracking the time-varying nature of the interference. 1 Introduction In this paper, we consider the problem of detecting, tracking and suppressing interference in a spread-spectrum communication system. The interference is assumed to consist of many unknown, independent and time-varying narrowband interferers. There are several advantages with spreading the spectrum of the signal th...

### Citations

8198 | Maximum likelihood from incomplete data via the EM algorithm
- Dempster, Laird, et al.
- 1977
(Show Context)
Citation Context ...rer is presented here. In order to simplify the notation for the reader, we here drop the superscript 'm' indicating the mth interferer. Interference Parameter Estimation The EM algorithm proposed in =-=[1-=-9] is used to obtain the maximum likelihood (ML) estimate of which is denoted here as ML . As a by-product of the E-step, conditional mean estimates of the state A of the mth interference is obtaine... |

4306 | A tutorial on Hidden Markov Models and selected applications in speech recognition
- Rabiner
(Show Context)
Citation Context ...ithm converges to a stationary value of the likelihood function. The superscript '(l)' indicates the iteration number of the EM algorithm. Some of the results presented in this section are derived in =-=[-=-21]. The EM algorithm is outlined in Fig. 2. First, the lth iteration of the E-step in the above mentioned EM algorithm is performed. The Expectation Step: 8 The aim in the E-step is to evaluate Q ; ... |

329 |
On the convergence properties of the EM algorithm
- Wu
- 1983
(Show Context)
Citation Context ... the maximum likelihood (ML) estimate of which is denoted here as ML . As a by-product of the E-step, conditional mean estimates of the state A of the mth interference is obtained. It is shown in [2=-=0-=-] that under mild regularity conditions, the sequence n (l) o for l = 1; 2; : : :, of the EM algorithm converges to a stationary value of the likelihood function. The superscript '(l)' indicates the ... |

161 |
Communication Systems
- Haykin
- 1987
(Show Context)
Citation Context ...=0 X m (kF s ; b) (11) for k = 0; : : : N 1; b = 1; : : : T . By choosing the sampling period T s and the number of data points N per batch, such that F s = F and N = M , we have the following result =-=[18-=-] X m (kF s ; b) = ( NT s A m b ; if m = k 0; otherwise k = 0; : : : N 1; b = 1; : : : T: (12) The DFT of the observation noise is given by W (kF s ; b) = T s bN 1 X n=bN N w n e j 2n N k ; k = 0; : :... |

134 |
Theory of spreadspectrum communications - a tutorial
- Pickholtz, Schilling, et al.
- 1982
(Show Context)
Citation Context ... is assumed to consist of many unknown, independent and time-varying narrowband interferers. There are several advantages with spreading the spectrum of the signal that you wish to transmit; see e.g. =-=[2, 3]-=-. One of the major advantages is the inherent ability to reject interfering signals whose bandwidths are small compared to that of the spread spectrum. However, the interference may be powerful enough... |

102 |
Acceleration of stochastic approximation by averaging
- Polyak, Juditsky
- 1992
(Show Context)
Citation Context ...35) K b denotes the step size at the bth time instant and is chosen as K b = 1 b : (36) 12 In order to speed the convergence, K b could instead be chosen as b l , where l is a positive, real constant =-=[-=-24]. I. The transition probabilities, . By using a dierential geometric approach suggested in [25], the formulae for updating the transition probabilities are given by the following set of equations ... |

43 |
Interference rejection techniques in spread spectrum communications
- Milstein
- 1988
(Show Context)
Citation Context ... is assumed to consist of many unknown, independent and time-varying narrowband interferers. There are several advantages with spreading the spectrum of the signal that you wish to transmit; see e.g. =-=[2, 3]-=-. One of the major advantages is the inherent ability to reject interfering signals whose bandwidths are small compared to that of the spread spectrum. However, the interference may be powerful enough... |

43 |
Spread spectrum CDMA systems for wireless communications,2nd ed
- Glisic, Vucetic
- 1997
(Show Context)
Citation Context ... under Grand N00014-94-1-0115, and in part by the New Jersey Center for Mobile Telecommunications. y Parts of this paper has been presented at [1] 1 has been studied extensively over the last decades =-=[2, 3, 4-=-]. Previous work in this area can be classied into frequency domain and time domain approaches. Time domain techniques for narrowband interference can be split into linear and non-linear methods. [2] ... |

25 |
Iterative and sequential algorithms for multisensor signal enhancement
- Weinstein, Oppenheim, et al.
- 1994
(Show Context)
Citation Context ...terference present at each frequency bin. The estimation and detection of the mth interferer is presented below. Interference Parameter Estimation The on-line stochastic gradient algorithm studied in =-=[22-=-, 23] is applied to the posed problem. The model parameters are updated according to the following iterative scheme b = b 1 + 1 1 b 1 K b I b ; (33) where is a predetermined real constant and I b... |

17 |
V.: Nonlinear techniques for interference suppression in spreadspectrum systems
- Vijayan, Poor
- 1990
(Show Context)
Citation Context ...band interference exist in the literature; e.g. in [8, 9], the interference is modeled as a Gaussian AR process, in [7] as a sum of sinusoids and in [6] as a pulsed RF tone. In 1991, Vijayan and Poor =-=[10]-=- suggested a non-linear technique for prediction of the narrowband interference signal that took into account the non-Gaussian distribution of the observation noise. This led to various nonlinear tech... |

16 | Narrowband interference suppression in spread spectrum CDMA
- Poor, Rusch
- 1994
(Show Context)
Citation Context ...inear and non-linear methods. [2] provides a discussion on parametric linear techniques for narrowband interference suppression. In these linear techniques, also known as estimator/subtracter methods =-=[5]-=-, a transversalslter is used to obtain estimates of the received signal based on previous samples and model assumptions. Theslter is implemented using single-sided taps (linear predictionslter) or dou... |

15 | Narrowband interference suppression in CDMA spread spectrum communications
- Rusch, Poor
- 1994
(Show Context)
Citation Context ...signal that took into account the non-Gaussian distribution of the observation noise. This led to various nonlinear techniques to combat narrowband interference suppression in spread spectrum systems =-=[5, 10, 11]-=-. The interference in [5, 10, 11] is modeled as a Gaussian AR process. For known interference statistics the interference is estimated using an approximate conditional mean (ACM)slter [12]. The ACMslt... |

11 |
Proakis “Adaptive algorithms for estimating and suppressing narrow band interference in PN spread spectrum systems
- Ketchum, G
- 1982
(Show Context)
Citation Context ...rpolationslter). Interpolation linearslters were found to give greater interference suppression. Adaptive schemes based on the least mean squares (LMS) algorithm and latticeslters have been developed =-=[6, 7, 8-=-]. Dierent models for narrowband interference exist in the literature; e.g. in [8, 9], the interference is modeled as a Gaussian AR process, in [7] as a sum of sinusoids and in [6] as a pulsed RF tone... |

8 |
Rejection of pulsed CW interference in PN spread-spectrum systems using complex adaptive filters
- Li, Milstein
- 1983
(Show Context)
Citation Context ...rpolationslter). Interpolation linearslters were found to give greater interference suppression. Adaptive schemes based on the least mean squares (LMS) algorithm and latticeslters have been developed =-=[6, 7, 8-=-]. Dierent models for narrowband interference exist in the literature; e.g. in [8, 9], the interference is modeled as a Gaussian AR process, in [7] as a sum of sinusoids and in [6] as a pulsed RF tone... |

7 |
Adaptive narrow-band interference rejection in a DS spread-spectrum intercept receiver using transform domain signal processing techniques
- Gevargiz
- 1989
(Show Context)
Citation Context ...e. Frequency domain techniques are usually non-parametric and require no prior knowledge of the characteristics of the interference. These algorithms are based on transform domainsltering; 2 see e.g. =-=[7, 15, 16]-=-. The key idea is that the received signal is estimated and used to designslters that attenuate the signal in the frequency range where the interference is dominant. [7] employs the fast Fourier trans... |

6 |
Closed-Form Analytical Results for the Rejection of Narrow-Band Interference
- Masry
- 1985
(Show Context)
Citation Context ...rpolationslter). Interpolation linearslters were found to give greater interference suppression. Adaptive schemes based on the least mean squares (LMS) algorithm and latticeslters have been developed =-=[6, 7, 8-=-]. Dierent models for narrowband interference exist in the literature; e.g. in [8, 9], the interference is modeled as a Gaussian AR process, in [7] as a sum of sinusoids and in [6] as a pulsed RF tone... |

2 |
Approximate non-gaussian with linear state and observation relations
- Masreliez
- 1975
(Show Context)
Citation Context ...tems [5, 10, 11]. The interference in [5, 10, 11] is modeled as a Gaussian AR process. For known interference statistics the interference is estimated using an approximate conditional mean (ACM)slter =-=[1-=-2]. The ACMslter is a modication of the Kalmanslter that deals with non-Gaussian distributions in the observation or the state process. For the specied assumptions on the observation process in [5, 10... |

2 |
Spread spectrum receiver using acoustic surface wave technology
- LB, Das
- 1977
(Show Context)
Citation Context ...e. Frequency domain techniques are usually non-parametric and require no prior knowledge of the characteristics of the interference. These algorithms are based on transform domainsltering; 2 see e.g. =-=[7, 15, 16]-=-. The key idea is that the received signal is estimated and used to designslters that attenuate the signal in the frequency range where the interference is dominant. [7] employs the fast Fourier trans... |

2 |
Suppression of narrowband jammers in a spread-spectrum receiver using transform-domain adaptive
- Saulnier
- 1992
(Show Context)
Citation Context ...rsalslter is designed for suppressing the interference. In [15, 16], real-time Fourier transformation using a surface acoustic wave (SAW) with a chirp impulse response built into the taps is used. In =-=[17-=-], the problem of suppressing many independent narrowband interferers that are present in dierent frequency bins is studied. Adaptivesltering in the frequency domain via the LMS algorithm is performed... |

2 |
Iterative and recursive estimators for hidden Markov errors{in{variables models
- Krishnamurthy, Logothetis
- 1996
(Show Context)
Citation Context ...terference present at each frequency bin. The estimation and detection of the mth interferer is presented below. Interference Parameter Estimation The on-line stochastic gradient algorithm studied in =-=[22-=-, 23] is applied to the posed problem. The model parameters are updated according to the following iterative scheme b = b 1 + 1 1 b 1 K b I b ; (33) where is a predetermined real constant and I b... |

1 |
Suppression of High-Density, Dynamic Narrowband Interference
- Carlemalm, Poor, et al.
- 1999
(Show Context)
Citation Context ...h was supported in part by the U.S. Oce of Naval Research under Grand N00014-94-1-0115, and in part by the New Jersey Center for Mobile Telecommunications. y Parts of this paper has been presented at =-=[1-=-] 1 has been studied extensively over the last decades [2, 3, 4]. Previous work in this area can be classied into frequency domain and time domain approaches. Time domain techniques for narrowband int... |

1 |
Adaptive nonlinear for narrowband interference suppression in spread spectrum CDMA", under revision for publication in
- Krishnamurthy, Logothetis
- 1996
(Show Context)
Citation Context ...s. For the specied assumptions on the observation process in [5, 10, 11], the ACMslter for interference estimation turns out to be a Kalman-type recursiveslter which includes some nonlinearities. In [=-=13]-=-, a new (suboptimal) nonlinearslter and parameter estimator for narrowband interference suppression in spread spectrum systems is presented. A cross-coupled hidden Markov model (HMM) and a Kalmanslter... |

1 |
On-Line Identi of Hidden Markov Models via Recursive Prediction Error Techniques
- Collings, Krishnamurthy, et al.
- 1994
(Show Context)
Citation Context ...er to speed the convergence, K b could instead be chosen as b l , where l is a positive, real constant [24]. I. The transition probabilities, . By using a dierential geometric approach suggested in [2=-=-=-5], the formulae for updating the transition probabilities are given by the following set of equations ij;b = g ij;b P 2 i;j=1 (g ij;b ) 2 ; (37) where ij;b denotes the transition probability estima... |