Project
Phase Retrieval and its Extension: Algorithms and Analysis
- Phase retrieval refers to the recovery of a signal given only the intensity measurements and has wide applicability. However, it requires minimization of a multivariate fourth-order polynomial, indicating that the computational cost is very high especially when the signal length is long. The aim of this research is to explore signal processing and optimization techniques to devise efficient algorithms for phase retrieval and its related problems. We start with utilizing the coordinate descent to retrieve the signal from its magnitude-square observations. Its main idea is to solve a single unknown at each iteration while all other variables are kept fixed, and thus only minimization of a univariate quartic polynomial is needed. Algorithm robustification particularly in the presence of outliers and extension to sparse signals, special linear mappings as well as phaseless matrices will then be investigated. Another challenging task is to produce the local/global convergence analysis of the proposed methods.
Near-Optimal Robust Beamforming: Algorithms and Analysis
- Beamforming refers to reception of the signal-of-interest (SOI) at a possibly known direction while suppression of the surrounding interferences and noise, with the use of spatially separated sensors via properly selecting their weights. It has been an important and fundamental task in numerous application areas including radar, sonar, communications, acoustics, astronomy, seismology and biomedical engineering. Modern beamformer design computes the array weights as a function of the received data according to an optimization criterion. Nevertheless, there are challenges that have not been solved. First, even the state-of-the-art robust beamformers cannot attain optimum performance in terms of signal-to-interference-plus-noise ratio, particularly in the practical scenarios when the SOI and interferences are non-Gaussian distributed. Second, many beamforming algorithms require high computational complexities, indicating that their applicability is limited. In this research, we aim to explore advanced signal processing and optimization techniques to devise robust beamformers with near-optimal performance as well as computational attractiveness.
Non-convex Optimization for Robust Sparse Recovery: Fast Algorithms and Theoretical Analysis
- Sparse recovery refers to extracting a high-dimensional vector with few nonzero entries from a small number of linear measurements. It has been a core topic in various areas of science and engineering including statistics, signal processing, machine learning, information theory, medical imaging and computer vision, because the real-world signals of interest often have a sparse representation in some basis. As an intensive field of research, however, there are obstacles need to be overcome in order to further enhance its practicality. One key issue is to solve large-scale problems in big data analytics where the number of variables is enormous, implying the need of computationally attractive solutions for sparse recovery. Another challenge is to recover the sparse signals from as few observations as possible. A representative example is in magnetic resonance imaging where significant scan time reduction means benefits for patients and health care economics. Furthermore, the development of most existing sparse recovery algorithms assumes that the measurement noise is Gaussian distributed. However, the occurrence of non-Gaussian impulsive noise is in fact common in many applications, and thus these standard solvers may not be able to provide reliable performance in such scenarios.
In this research, via exploring advanced signal processing and optimization methods and theories, we propose to utilize possibly non-convex sparsity-inducing and noise-resistant functions in devising efficient and robust algorithms to recover sparse signals in non-Gaussian noise environment with minimum observations. Their fast and distributed realizations will be produced. Moreover, theoretical performance metrics of the developed algorithms, namely, local and global convergence as well as conditions of exact recovery, which are very challenging to analyze in a non-convex setting but are really vital, will be studied. We will also employ the non-convex based sparse recovery methodology to important applications including spectral estimation, source localization, image denoising, magnetic resonance imaging and social network analysis.
Recovery of Low-Rank Matrices/Tensors in Impulsive Noise
- Low-rank matrices and tensors frequently arise in many areas of science and engineering including applied mathematics, data mining, chemometrics, bioinformatics, medical imaging, machine learning, computer vision and graphics. The core objective of low-rank matrix/tensor recovery is to retrieve them from limited and/or noisy linear observations by incorporating the small rank constraint. Conventional techniques for matrix/tensor recovery often rely on the Gaussian assumption of the measurements and their derivation is based on the l2-space. As a result, these standard solvers may fail to work properly when the observations contain outliers. In this research, we aim to explore advanced signal processing techniques in the lp-space, where p is less than 2, to achieve outlier-resistant matrix/tensor recovery. The concepts of lp-norm minimization and best rank-one matrix/tensor fitting in the lp-space will be exploited to derive robust convex/non-convex optimization and greedy pursuit algorithms. Furthermore, performance metrics of the developed algorithms will be studied.
Robust Sparse Signal Recovery in lp-Space
- A sparse signal can be modelled as a linear combination of very few elements drawn from a fixed collection. Retrieving the signal vector with only a small fraction of nonzero entries in some basis from the observations is referred to as sparse signal recovery, which corresponds to many real-world applications in science and engineering. Nevertheless, conventional techniques for signal recovery often rely on the Gaussian assumption of the measurements and their derivation is based on the l2-space. As a result, these standard solvers may fail to work properly when the observations contain outliers. In this research, we aim to explore advanced signal processing techniques in the lp-space, where p is less than 2, to achieve outlier-resistant signal recovery. The concept of lp-correlation will be exploited to accurately determine the similarity between the elements and the target signal embedded in impulsive noise.
Outlier-Resistant Parameter Estimation in lp-Space
- The classical approach for solving many problems encountered in science and engineering relies on the Gaussian assumption of the data. In spite of providing theoretical and computational convenience, it is generally understood that the validity of the Gaussian assumption is at best approximate in reality. In fact, the occurrence of non-Gaussian impulsive noise has been reported in many fields such as radar, sonar, wireless communications, biomedical engineering and image processing. Even applying the optimal methods based on the Gaussian assumption to these scenarios will result in poor performance. In this project, we tackle the topic of estimating the parameters of interest from observations which contain outliers. The lp-space, where p is less than 2, will be exploited to achieve outlier-resistant parameter estimation. Computationally efficient and fast converging lp-norm minimization algorithms will be devised. The very challenging case of 0 < p < 1, which correspond to non-convex optimization, will also be addressed.
Network Localization Using Received Signal Strength Measurements
- Knowing the positions of targets of interest in a wireless network is essential in numerous application areas such as emergency assistance, personal monitoring, fleet management, asset tracking, location-based advertisement and billing, as well as travel services. Among various localization strategies, received signal strength (RSS) based positioning systems, which utilize the radio signal attenuation to provide range information, are attractive in terms of cost and simplicity. It is because RSS measurements are readily available in the existing infrastructures such as Wi-Fi, ZigBee and cellular networks. In this research, we propose to utilize advanced statistical signal processing theories and techniques to develop RSS based positioning algorithms which can attain high localization accuracy and computational attractiveness, and are less sensitive to unknown or uncertain environments. The performance of the devised schemes will also be analyzed and evaluated in different wireless network scenarios.
Advanced Signal Processing for Target Enumeration and Localization in Multiple-Input Multiple-Output Radar
- Radar is an acronym for radio detection and ranging system. Multiple-input multiple-output (MIMO) radar is an emerging technology which uses a number of spatially separated transmitters and receivers to simultaneously transmit radio waves and to receive the reflected signals for joint processing, respectively, in order to achieve the important tasks of target detection, positioning and imaging. Unlike the conventional phased-array radar which can only transmit scaled versions of a single waveform, the transmit antenna arrays of the MIMO radar are capable of emitting independent and orthogonal signals. The waveform diversity in the MIMO radar paradigm has been shown to offer better identifiability (that is, more targets can be uniquely identified) and higher spatial resolution (that is, more closely-spaced targets can be resolved) over the phased-array counterpart, which leads to significant improvements in signal detection and parameter estimation performance. Nevertheless, many research challenges need to be tackled to advance MIMO radar from concept to reality. For example, a key difficulty is to achieve optimum detection and estimation performance particularly when the signal-to-noise ratio is low or the number of snapshots is small. Furthermore, processing of the multi-dimensional radar data corresponds to enormous computations and thus algorithm complexity is a main concern.
The aim of this research is to devise advanced signal processing approaches for determining the number of targets and finding their positions in an accurate and computationally efficient manner by exploiting the special structure inherent in the MIMO radar signals. In particular, we will utilize random matrix theory for target detection and base on the subspace approach as well as tensor algebra for localization algorithm development. The performance of the devised methods will also be analyzed and evaluated for various MIMO radar application models.
Advanced Signal Processing for Multidimensional Harmonic Retrieval
- The problem of harmonic retrieval (HR) is to extract the parameters from noisy sinusoids, which has been an important topic in science and engineering because many real-world signals can be well described by the sinusoidal model. Although one-dimensional (1-D) HR is the most common, multidimensional HR in fact has numerous applications such as wireless communication channel estimation, nuclear magnetic resonance (NMR) spectroscopy and multiple-input multiple-output (MIMO) radar imaging. For example, the measurements in NMR spectroscopy which is a powerful technique for protein research in food and nutritional industries, can be modeled as a sum of multidimensional damped sinusoids where the frequencies and damping factors are crucial to determining the protein structures. Moreover, the sinusoidal parameters of the MIMO radar data contain the position information of multiple targets of interest.
Analogous to 1-D HR, the key step in the multidimensional scenarios is to find the damping factor and frequency parameters because they are nonlinear functions in the observed signals. However, multidimensional HR is much more challenging than the 1-D counterpart. First, it is difficult to align the parameters at all dimensions particularly when there are identical frequencies in at least one dimension. Second, processing of multidimensional data corresponds to enormous computations and thus algorithm complexity is a main concern. Third, even the data dimension is more than two, the multidimensional signals are stored in matrices by means of stacking operations for manipulation in most of the existing approaches, implying that the harmonic structure may not be fully utilized in the estimation procedure.
The aim of this research is to develop advanced multidimensional HR approaches that are efficient in both aspects of accuracy and complexity. To fulfill the challenging demands, we propose to effectively exploit tensor algebra which perfectly aligns with multidimensional data as well as principal singular vectors and values determined from the observed signals for optimal parameter estimation. Moreover, we will extend the proposed methodology for estimating the number of sources in case it is not known a priori. The performance of the devised multidimensional HR schemes will also be analyzed and evaluated in different applications.
Novel Subspace Approach for Efficient Sinusoidal Parameter Estimation
- Parameter estimation of sinusoids from a finite number of noisy discrete-time measurements is an important topic in science and engineering because numerous real-world signals can be well described by the sinusoidal model. Although many spectral estimation methods are available in the literature, there is generally a tradeoff between estimation performance and computational requirement. The aim of this research is to develop subspace-based sinusoidal parameter estimation approaches that are efficient in both aspects of accuracy and complexity. To fulfill these challenging demands, we suggest to effectively utilize the principal singular vectors determined from the observed signals. Moreover, we will extend the proposed methodology for estimating the number of sinusoids in case it is unknown. Performance analysis of the devised spectral estimators will also be derived.
Convex Optimization for Frequency Estimation and Related Problems
- Estimation of the frequencies of sinusoidal signals from a finite number of noisy discrete-time measurements has been an active research area to date because of its wide applications in science and engineering. Although the maximum likelihood estimator, nonlinear least squares technique and periodogram can provide optimum frequency estimation performance, the cost functions of all these methods are multimodal and two steps are typically involved in the estimation procedure: suboptimal initial parameter estimates are first obtained and then a refinement is made through an iterative optimization of the cost functions. As a result, sufficiently accurate initial estimates are critical to achieve the globally optimum solutions. In this research, we propose to approximate these estimators using convex optimization technique so that high-performance global solutions can be obtained. Related signal processing problems, namely, parameter estimation of polynomial-phase signals, system identification and polynomial root finding, will also be investigated in the convex optimization framework.
Localization of Tactile Interactions for Tangible Acoustic Interface Development
- Human-computer interaction devices are classifiable as tangible or touchable and intangible or non-touchable interfaces. Typical examples of the former are keyboard and mouse while the latter includes graphical display and loudspeakers. Nevertheless, existing tangible interfaces require users to be in certain locations when interacting with computers. To enhance user mobility, one innovative and exciting idea is to convert any touchable objects such as tables, walls and windows into natural and unrestricted interactive surfaces, which are referred to as tangible acoustic interfaces (TAIs). The underlying principle is to utilize the acoustic vibration produced by the user contact with the objects and thus localization of the tactile interactions is a fundamental and crucial issue for the TAI development.
A common positioning strategy is to use time-difference-of-arrival (TDOA) measurements obtained from acoustic sensors to derive the location estimate. However, TDOA estimation in TAIs is a very difficult task because acoustic propagation in solid media is well-known to be a very complex physical phenomenon to characterize and model. In this research, we aim to devise custom-made and reliable TDOA estimation and positioning algorithms for TAI applications.
Noise and Echo Cancellation DSP Firmware Library for Automotive Applications
- The demand for devices that utilize digital speech processing is constantly growing. Nowadays in many countries, the use of phone in cars is only allowed with hands-free equipped mobile phone. These hands-free facilities have loudspeakers and microphone installed in the car. It aims to allow communication in noisy and enclosed environment without using close-talking microphones, and of course, operated without hands. However, as the microphone is positioned some distance from the talker, this will increase the loss in the transmission loop as well as increase the level of ambient noise and the echo from the loudspeaker signal.
The aim of this project is to restore the comfort of a face-to-face conversation over an in-car hands-free phone. Based on the limitations of the in-car environment, the project will develop a digital speech processing algorithm which will be simulated and implemented in real-life applications.
Unified Power Management for Collaborative Wireless Sensor Networks
- Recent years have seen the deployment of wireless sensor networks (WSNs) in a variety of applications such as asset management, environment monitoring, health care, building automation, and surveillance. The success of these applications hinges on the following: (i) nodes must effectively collaborate with each other to achieve complex sensing and communication tasks, as each node has very limited capability; and (ii) due to limited power supplies (e.g., batteries), network power consumption must be minimized to achieve the long lifetime required by applications.
This project investigates novel approaches that can minimize the power consumption of WSNs while maintaining desired network performance. In contrast to previous studies that addressed sensing and communication requirements separately, the researchers integrate them into unified power management models. They will investigate two power management problems based on stochastic and collaborative sensing models: (i) minimum-power network topology that enables collaborating nodes to achieve desired data fusion performance; and (ii) delay-efficient sleep schedules that minimize nodes' idle power consumption in data fusion. Localized network protocols will be developed by exploring the fundamental relationship between sensing and communication of WSNs. The results are expected to enable a wide range of WSN applications to achieve assured performance throughout an extended network lifetime.
Time Delay Estimation Based Speaker Localization
- In applications such as video conferencing, automatic scene analysis and security monitoring, there is a need to determine the positions of active talkers. A simple and effective technique for automatic speaker location is to use an array of spatially separated microphones, and two steps are involved as follows. At first, a set of differences in arrival times of the acoustic source signal received at multiple pairs of microphones is measured. This time difference information is then employed to estimate the speaker position with the known microphone array geometry.
In this research, we propose to develop an accurate time difference estimation algorithm particularly for the speech signals received at a microphone array, via exploring the speech properties. Furthermore, a reliable and efficient location method will be devised based on the time difference measurements, subject to the constraints derived from the available prior knowledge such as admissible speaker range and other cues.
Node Localization for Wireless Sensor Networks: Algorithm Development and Performance Analysis
- Recent technological advances in wireless communications and microsystem integration have enabled the development of small, inexpensive, low-power sensor nodes which are able to collect surrounding data, perform small-scale computations and communicate among their neighbors. These wirelessly connected nodes, when working in a collaborative manner, have great potential in numerous remote monitoring and control applications such as asset management, habitat monitoring, health caring, building automation, battlefield surveillance as well as environment observation and forecasting. Since sensor nodes are often arbitrarily placed with their positions being unknown, sensor positioning is a fundamental and crucial issue for the wireless sensor network (WSN) operation and management. Time-of-arrival, received signal strength and/or angle-of-arrival measurements between sensor pairs are commonly used to estimate the node positions. Nevertheless, a major requirement for the WSN localization problem is to provide location estimates which are accurate via efficient utilization of the available pair-wise measurements and low-cost because of the limited processing and communication capabilities resulted from the nodes' battery constraint. Furthermore, the WSN connectivity is subject to changes due to node addition and failure, nonstationary environment as well as sensor mobility. It is also important to detect and discard non-line-of-sight measurements, which arise when direct signal transmission paths between sensor pairs are blocked, because they will introduce large errors in position estimation. In this research, the above challenges will be tackled and we propose to utilize the statistical signal processing theory and techniques to derive accurate, robust and computationally attractive WSN positioning algorithms. The bias and variance performance measures of the developed algorithms will also be produced.
A Study on Particle Filtering for Sinusoidal Parameter Estimation
- Estimation of the frequencies, amplitudes and/or phases of sinusoidal signals from a finite number of noisy discrete-time measurements has been an active research area to date because of its wide applications in science and engineering. Although numerous sinusoidal parameter estimation methods have been suggested in the literature, most of them address quasi-stationary parameters and/or additive Gaussian noise environments. Nevertheless, the sinusoidal parameters are time-varying as well as the noise is non-Gaussian and/or multiplicative in many real-life application areas such as audio signal processing and digital communications. In this research, particle filtering which is an advanced and promising approach for tracking applications in nonlinear and/or non-Gaussian processes, will be exploited in developing accurate yet robust algorithms for sinusoidal parameter estimation in these challenging scenarios. The performance of the proposed algorithms will be evaluated using computer simulations as well as real data from audio signal processing applications.
Statistical Signal Processing for Subsurface Sensing: Fundamental Investigation and Applications
- The problem of detecting anomalous objects that are underground, underwater or embedded man-made structures has diverse application areas such as transportation infrastructure assessment, utilities mapping, building diagnosis, archaeology and humanitarian demining. Ground penetrating radar (GPR) is a suitable tool for subsurface object detection because it enables fast and non-destructive subsurface sensing, and is relatively inexpensive. The basic operating principle of GPR is that it emits electromagnetic wave to the inspection surface and receives the returned signal from which the decision of whether there is an anomaly is made. Nevertheless, GPR signal is very susceptible to surface bounce and clutter interference, and signal processing must be applied in order to improve detection and reduce false alarms. In fact, it is a challenging task to differentiate between the returned signals correspond to the background clutter interference and the interested anomaly, particularly when the background environment and/or target anomaly signal are nonstationary due to nonuniformities inside the surface, as well as human factors such as variations in sweeping speed and radar-to-surface distance if hand held GPR is employed. In this research, we aim to utilize statistical signal processing techniques, particularly parameter estimation and signal detection, to tackle the subsurface sensing problem. It is expected that accurate yet robust signal models of the returned GPR data can be developed via exploiting estimation theory while detectors that maximize the detection rate and minimize the probability of false alarm can be derived with the use of detection theory. Practical experiments for potential applications of flaw detection and water seepage diagnosis in buildings, crack detection in bridges and leak detection in underground water pipes, will be conducted to validate the performance of the proposed methods.
Advanced Signal Processing for Automotive Applications
- In the last decade, the impact of information technology on the design and performance of road-vehicles (including cars, trucks, coaches and agricultural tractors) has been enormous. A major ingredient of the genuine innovation in newly designed cars nowadays is driven by digital control and digital signal processing.
The focus of this project is to tackle three highly innovative automotive applications where high-performance signal processing plays a fundamental role, namely:
(i) Estimation of road-tire contact forces using sensors embedded in tires, which are referred to as smart tires;
(ii) Active reduction of torque-pulses in four-stroke internal-combustion engines;
(iii) Active acoustic noise reduction in cars.
The common and crucial element in solving these three problems is to accurately and rapidly track the sinusoidal signals embedded in unknown broadband noise. It is expected that their successful solutions will contribute to the improvement of road-vehicle system performance, increase of passenger satisfaction and safety as well as direct benefits to environmental protection. In this research, advanced signal processing and parameter estimation techniques will be developed for estimation, rejection and tracking of the periodic components in the real data collected from the three automotive applications.
Multiple-Input/Multiple-Output Channel Estimation with Infinite Impulse Response Filters
- Space-time processing (STP) techniques are a key way to enhance the capability of mobile communication services in the long term, and are currently regarded by many within the wireless communications industry as a core system component in future-generation mobile networks. The key idea of the STP system is to use an antenna array to process signal samples both in time and space. The extra spatial dimension in STP enables interference cancellation and increases capacity in a way that is not possible with the commonly used temporal-only processing systems. For an STP system where signals of multiple users are received at a base station employing multiple antennas, such a signal model is known as a multiple-input/multiple-output (MIMO) model. To recover the transmitted information of each user, the MIMO propagation channel should be accurately identified. In the literature, the MIMO channel is usually modeled as a set of finite impulse response (FIR) filters. The advantages of using FIR filters are that they are always stable and have global convergence property. However, they typically require much larger filter coefficients than the general filtering model - infinite impulse response (IIR) filters, which should model the MIMO channel in a better sense. In this research, we aim to develop stable and accurate MIMO channel estimators based on IIR filtering.
Wireless Location of a Mobile Station
- Accurate localization of mobile phones is receiving considerable attention in the field of wireless communications. The primary motivation is to determine the position of a mobile station (MS) whose caller is in an emergency situation but unable to describe its location. In addition to emergency management, mobile position information would also be useful in applications such as car navigation, fleet operations and locating lost or stolen mobiles. Wireless location systems usually require three or more base stations (BSs) to intercept a MS signal. The common approach uses the time difference of arrival (TDOA) measurements of the signal received at the BSs to derive the location estimate of the MS. Unfortunately, the presence of multipath propagation, which is always present in wireless communications, has made the task of TDOA estimation very difficult. Another source of estimation error occurs when the direct or line-of-sight (LOS) path between the MS and BS is blocked by buildings or mountains. Then it becomes necessary to perform TDOA estimation with non-line-of sight (NLOS) propagation only. Furthermore, the existence of phase noise, which is invariably produced from imperfect oscillators employed in the modulation and demodulation of the received signals, can lead to unreliable TDOA measurements. In this research, we aim to mitigate these impairments and combine other possible sources of information in order to increase the location accuracy.
Unbiased Instantaneous Frequency Estimation using Linear Prediction
- Instantaneous frequency (IF) is used to describe a signal's frequency which varies with time. Estimation of the IF has many important applications in seismic, radar, sonar, communications, etc., where the signals involved are nonstationary. Examples include demodulation of a frequency-shift keying (FSK) signal in digital communications and coding of a speech waveform in multimedia applications. Discrete-time energy separation algorithm (DESA) and linear prediction method are two common approaches for IF estimation. These techniques usually require 3 to 5 discrete-time samples of the signal for determining the IF and thus real-time computation is allowed. Another advantage of them is that they provide closed-form formulae for IF estimation so that the IF estimates are directly obtained from the received samples. However, it is shown that the IF estimates provided by these two approaches are biased in the presence of noise. In this project, we aim to devise an unbiased IF estimation method under noisy environments, by using a modified linear prediction technique.
Low-Complexity Blind Space-Time Processing for Wireless Communications
- Space-time processing (STP) techniques are a key way to enhance the capability of mobile communication services in the long term, and are currently regarded by many within the wireless communications industry as a core system component in future-generation mobile networks. Basically, a STP system uses an adaptive antenna array to process signal samples both in time and space. The extra spatial dimension in STP enables interference cancellation in a way that is not possible with time-only processing (TOP) systems. In this research, we aim to develop efficient, low-complexity, blind STP methods for reliable detection of multiple mobile signals, which is complicated by many interferences and environmental noise sources.
Joint AOA-TDOA-FDOA Estimation for Mobile Phone Location
- The problem of locating mobile phones in a wireless communication system has received significant attention over the past few years. This has been motivated by a recent order issued by the Federal Communications Commission (FCC) that mandates all wireless service providers to locate an emergency 911 caller within a high accuracy by 2001. Apart from emergency management, mobile station (MS) location finding also has potential applications such as mobile yellow pages, cellular planning and intelligent transport systems. Wireless location is commonly achieved by measuring the angle-of-arrival (AOA) of the MS signal at several base stations (BSs), or by estimating the time-difference-of-arrival (TDOA) or frequency-difference-of-arrival (FDOA) between MS signals received at spatially separated BSs. In this research, we propose to jointly estimate the AOA, TDOA and FDOA from the received MS signals so that optimum parameter estimation can be attained. Reliable location solution based on the joint use of these parameter estimates will also be derived.
Robust Mixed Norm Adaptive Filter Algorithms with Time-Varying Step Sizes
- The least-mean-square (LMS) algorithm has been successfully applied in communications, control, sonar, and biomedical engineering. It is well known that the LMS algorithm provides satisfactory performance in the presence of Gaussian distributed measurement noise. However, this method degrades and may be even unstable for other measurement noises such as impulsive noise, uniformly distributed random process and sinusoidal interference. In this project, mixed-norm adaptive algorithms based on minimizing an appropriate combination of norms, in order to give promising performance in different kinds of noise, will be developed and analyzed. Design of optimum time-varying step sizes for the algorithms will also be investigated.
Adaptive Algorithms for Sinusoidal Interference Cancellation
- Elimination of sinusoidal interference from an observed signal has found applications in many areas such as communications, control and biomedical engineering. A typical example is to remove unwanted 50Hz interference due to magnetic induction and displacement current from power line in electrocardiography. Based on adaptive noise canceling approach, a computationally efficient algorithm which provides direct amplitude and phase measurements for eliminating the interfering signal, assuming that the frequencies of the sinusoidal interference are known a priori, is developed. Performance measures of the algorithm, viz convergence behaviors and mean square errors of the system parameters as well as improvement of signal-to-noise ratio, are analyzed.