Adaptive Signal Processing Techniques Research Articles

Temporal pattern and spectral complexity as stimulus parameters for eliciting a cardiac orienting reflex in human fetuses.

The purpose of this study was to determine whether temporal pattern and/or spectral complexity were important stimulus parameters for eliciting a cardiac orienting reflex (OR) in low-risk human fetuses. Each of 28 term fetuses was exposed to four sounds formed from the four different combinations of temporal pattern (pulsed, continuous) and spectral complexity (sine wave, /â/). The fetal cardiac electrical signal was captured transabdominally at a rate of 1024 Hz, and fetal R-waves were extracted by using adaptive signal-processing techniques. We found that pulsed sounds elicited a significantly greater decrease in heart rate (HR) than did continuous sounds. However, the HR response was relatively unaffected by spectral complexity. For the pure tone and the phoneme used in this study, our results indicate that temporal characteristics were more effective at eliciting a cardiac OR in human fetuses than was spectral complexity.

Human fetuses have nonlinear cardiac dynamics.

Approximate entropy (ApEn) is a statistic that quantifies regularity in time series data, and this parameter has several features that make it attractive for analyzing physiological systems. In this study, ApEn was used to detect nonlinearities in the heart rate (HR) patterns of 12 low-risk human fetuses between 38 and 40 wk of gestation. The fetal cardiac electrical signal was sampled at a rate of 1,024 Hz by using Ag-AgCl electrodes positioned across the mother's abdomen, and fetal R waves were extracted by using adaptive signal processing techniques. To test for nonlinearity, ApEn for the original HR time series was compared with ApEn for three dynamic models: temporally uncorrelated noise, linearly correlated noise, and linearly correlated noise with nonlinear distortion. Each model had the same mean and SD in HR as the original time series, and one model also preserved the Fourier power spectrum. We estimated that noise accounted for 17.2-44.5% of the total between-fetus variance in ApEn. Nevertheless, ApEn for the original time series data still differed significantly from ApEn for the three dynamic models for both group comparisons and individual fetuses. We concluded that the HR time series, in low-risk human fetuses, could not be modeled as temporally uncorrelated noise, linearly correlated noise, or static filtering of linearly correlated noise.

Application of adaptive filter to active noise control

A modified LMS algorithm is suggested after a mathematical model for sound-cancelling, on the basis of the sound wave interference principle, is set up. Then a new adaptive active noise control up to 8 dB within the frequency range of 100 to 800 Hz can be achieved by way of using adaptive digital signal processing technique. In addition to this, there exits an attenuation of 11 dB on 500 Hz or so. The experiments about a pink noise, a broadband noise and a noise taken from a tank were earned out. They shows that the adaptive noise control technique is very effective and valid.

Application of nonlinear adaptive signal processing techniques to blind source separation and interference suppression

In active noise and vibration control it is important to identify independent noise generation mechanism and sources. In all large physical systems only a mixture of noise sources can be observed. The mixture model can be either linear or nonlinear. Recently there has been an explosion of significant research literature on blind deconvolution. In blind deconvolution, the actual mixture model and source signals are not known; however, in the theoretical development it is assumed that sources are statistically independent. In practice, good results have been obtained even if sources are not independent. If the mixture model is linear, statistically independent noise sources can be separated by blind deconvolution techniques that are based on the theory of the independent component analysis (ICA). We implement ICA by a cascade of the linear adaptive algorithm that is based on singular value decomposition (SVD) and a nonlinear algorithm that is based on a radial-basis function (RBF) neural network. A RBF neural network can also approximate the inverse of the nonlinear mixture model. We use a fast RBF updating algorithm for adaptation of the nonlinear network. Theoretical results are illustrated by computer simulations. Results are compared with linear adaptive signal processing algorithms. [This research has been supported by ONR, Dr. Kam Ng, Program Officer.]

The heart rate deceleratory response in low-risk human fetuses: effect of stimulus intensity on response topography.

The purpose of this study was to determine the effect of stimulus intensity on heart rate response in 18 low-risk human fetuses between 37 and 40 weeks gestation. Each fetus was stimulated in quiet sleep with a 30-s voice sound at intensities of 80 dB and 90 dB. The fetal cardiac electrical signal was captured transabdominally at a rate of 1024 Hz and fetal R-waves were extracted using adaptive signal-processing techniques. We found that fetuses generally exhibited a 5- to 10-s decrease in heart rate following stimulus onset at an intensity of 80 dB. The response pattern changed from deceleratory to acceleratory when stimulus intensity was increased to 90 dB. Our findings suggest that a heart rate deceleration at low-stimulus intensity may be a component of the orienting reflex in the human fetus.

Adaptive microphone arrays for speech enhancement in coherent and incoherent noise fields

To enhance the quality of speech in a hands–free telecommunication environment, microphone arrays consisting of multiple microphones can be used for sound pick–up. To circumvent the limitations associated with small array apertures adaptive signal processing techniques can be used. For extreme situations of a free (coherent) and a diffuse (incoherent) noise field, acceptable solutions for the noise reduction problem exist: In coherent noise fields, the adaptive beamforming techniques yield a high-noise reduction performance, but loose their efficiency in incoherent noise fields. On the other hand, for incoherent noise fields microphone arrays with adaptive post–filtering yield good performance; but in a coherent noise environment this method may cause unacceptable distortions in the output signal. In this contribution the advantages and disadvantages of these two techniques for noise reduction are discussed. A technique is presented which operates independently of the correlation properties of the noise fields, i.e., this method is able to reduce coherent and incoherent noise simultaneously. Some results are given using objective quality measures as well as subjective listening tests. The methods are also compared by using the recognition rate of a speaker-independent isolated word recognition system.

Improvement on active filter performance by employing an adaptive signal processing technique.

This paper proposes a simple and practical method for improving performance of DSP-controlled current source active power filters (AF). The transfer function of a DSP-controlled AF has a phase-lag nature because there is a time lag in the control output due to the computation time required by DSP, and the converter used for AF has first order lag characteristic. As AF has to generate high frequency compensation currents, the phase-lag characteristic causes a serious deterioration in the filter performance. In the proposed system, an adaptive digital filter (ADF) is used as a series compensator of AF and it compensates for the phase-lag characteristic of AF. It was confirmed by experimental results that a simple two-tap ADF is very effective to improve AF compensation performance.

Evaluation of Basal Autonomic Status in Normal Human Term Fetuses by Heart Rate Variability Spectral Analysis: Preliminary Report

Measures of autonomic nervous system functioning have received considerable attention because of the recognition that differences in sympathetic-parasympathetic balance may account for differences in neurobehavioral development. The purpose of this study was to determine if human fetuses could be categorized as either sympathetic-dominated or parasympathetic-dominated based on an analysis of the resting heart rate variability power distribution. The study population consisted of 17 normal human fetuses in quiet sleep between 36 and 40 weeks gestation. The fetal electrocardiogram was captured transabdominally in 3-min blocks at a rate of 833 Hz and fetal R-waves were extracted from the raw data using adaptive signal processing techniques. The fetal R-wave interbeat intervals (IBls) were converted to equally spaced, time-based data, and linear detrending was accomplished by subtracting the mean heart period from each weighted IBI. Total power (TP, 0.0-2.5 Hz) was divided into respiratory sinus arrhythmia (R...

Vagal tone during quiet sleep in normal human term fetuses.

The purpose of this paper was to calculate vagal tone (V) for 17 normal human fetuses in quiet sleep (QS) between 36 and 40 weeks gestation. The fetal cardiac electrical signal was captured transabdominally in 3-min blocks at a rate of 833 times per second and fetal R-waves were extracted using adaptive signal processing techniques. Fetal R-wave interbeat intervals were converted to equally spaced, time-based data, and the low-frequency component was removed using a 21-point third-order moving polynomial. The parameter V was calculated by taking the natural logarithm of the sum of the power densities between 0.3 Hz and 1.3 Hz. We found that fetal breathing was associated with an approximately 25% increase in V as compared to nonbreathing, 3.33 +/- 0.48 versus 2.57 +/- 0.47, p < 0.0001. Furthermore, there was a significant linear relationship between the mean single-fetus V during spontaneous respiration and the mean single-fetus V during normally occurring apneic periods, r = 0.772, p < 0.002. We conclude that respiratory activity is associated with a significant increase in vagal tone for normal human fetuses in QS.

Adaptive Compensation of Piezoelectric Sensoriactuators

An adaptive filter is used to estimate the feedthrough capacitance of a piezoelectric sensoriactuator in this work. The mechanical response of the piezostructure is resolved from the electrical response of the piezoelectric device through standard adaptive signal processing tech niques. Two common adaptive algorithms are reviewed for the given application: the LMS and the RLS. For spectrally white inputs the adaptation of the digital compensator yields a filter output which is proportional to the electrical response of the piezoelectric device. Thus, the remaining elec trical signal consists of the charge due to the mechanical response of the piezostructure. The adap tive filter converges to a combination of the feedthrough capacitance and the real portion of the mechanical piezostructure response, and the filter error is the quadrature component of the mechanical response. Preliminary results from the theoretical analysis and numerical simulations indicate that, under certain conditions, adaptive signal processing can be employed to realize an adaptive, sensoriactuator for simultaneous strain actuation and sensing.

Spectral analysis of heart rate variability during quiet sleep in normal human fetuses between 36 and 40 weeks of gestation

Respiratory sinus arrhythmia (RSA) is a clinical manifestation of the parasympathetic nervous system which can be identified in the high-frequency region of the heart rate variability (HRV) power spectrum. The purpose of this study was to determine the relative contribution of RSA to overall HRV for human fetuses in quiet sleep. The study population consisted of 13 normal human fetuses between 36 and 40 weeks of gestation for whom data were collected during spontaneous breathing and normally occurring apneic periods. Fetal breathing was monitored continuously using real-time sonography. The fetal electrocardiogram was captured transabdominally in 3-min blocks at a rate of 833 Hz and fetal R-waves were extracted from the raw signal using adaptive signal processing techniques. Fetal behavioral state was determined at the beginning and end of each 3-min data collection period. The fetal R-wave interbeat intervals (IBIs) were converted to equally-spaced, time-based data, and linear detrending of the time series was accomplished by subtracting the mean heart period from each weighted IBI. Total power (TP, 0.0–2.5 Hz) was divided into RSA (0.4–1.0 Hz), high-frequency (HF, 0.2–2.5 Hz), low-frequency (LF, 0.04–0.2 Hz), and very-low-frequency (VLF 0.0–0.04 Hz) regions, and the power densities were summed to determine the absolute power for each frequency component. A total of 81 3 min blocks (mean per subject 6.3, range 2–14) were available for analysis. Eleven (85%) of 13 fetuses demonstrated at HF peak during fetal breathing, and RSA accounted for approximately 20% of the TP. Although HF amplitude was reduced in the absence of fetal respiratory activity, RSA did not invariably disappear during periods of apnea, as the HF peak was recognized in seven (54%) fetuses when no respiratory activity was observed. Total power was approximately evenly distributed between the VLF, LF, and HF regions, and this distribution was independent of whether the fetus was breathing or not. However, the absolute power was significantly greater ( P < 0.05) during breathing as compared to non-breathing for TP and for the VLF, LF, HF, and RSA components. We conclude that RSA is common in normal human fetuses and is a significant factor contributing to HRV.

Adaptive compensation of piezoelectric sensoriactuators

An adaptive filter is used to estimate the feedthrough capacitance of a piezoelectric sensoriactuator in this work. The mechanical response of the piezostructure is resolved from the electrical response of the piezoelectric device through standard adaptive signal processing techniques. Two common adaptive algorithms are reviewed for the given application: the LMS and the RLS. For spectrally white inputs the adaptation of the digital compensator yields a filter output which is proportional to the electrical response of the piezoelectric device. Thus the remaining electrical signal consists of the charge due to the mechanical response of the piezostructure. An additional adaptive filter must be employed if a harmonic input is present since the harmonic input yields a bias term on the adaptive filter equal to the real part of the piezostructure response. Preliminary results from the theoretical analysis and numerical simulations indicate that adaptive signal processing can be employed to realize an adaptive, sensoriactuator for simultaneous strain actuation and sensing.

A robust algorithm for untrained adaptive signal processing

A new adaptive signal processing technique is described that does not require a training or reference signal for adaptation. The approach taken is to minimize variations in the complex envelope of an adaptive processor output while subjecting the processor coefficients to a set of linear constraints. This Linearly-Constrained Constant-Modulus (LCCM) method is motivated by the drawbacks associated with two existing untrained algorithms in certain signal environments. The problems specifically addressed by LCCM are the signal cancellation problem of Linearly-Constrained Power Minimization (LCPM) and the signal ambiguity problem of the Constant-Modulus Algorithm (CMA). A complete adaptive implementation of the LCCM algorithm is presented including its stability and convergence properties. In addition, adaptive performance of LCCM is demonstrated in a comparison with CMA and LCPM.

Dual mode adaptive signal processing

Adaptive signal processing using the least mean square (LMS) algorithm has been utilized in several different applications. The use of the adaptive echo canceller and the adaptive channel equalizer on many communications media has extended the rate of digital communication to near the theoretical channel capacity. Most applications of adaptive signal processing perform the task of system identification. Traditionally, this task is often performed by some predictor-corrector algorithm, with the optimal being the Kalman filter. In this respect, the LMS algorithm can be seen as an approximate limiting Kalman filter. On the other hand, another task of the traditional Kalman filter is signal smoothing. The signal smoothing task can also be performed by something close to the LMS algorithm as an approximate limiting Kalman algorithm. In this paper, the feasibility of using the LMS algorithm for signal smoothing is first explored. Then an architecture is proposed where the LMS algorithm is used for both signal smoothing and system identification. Finally, a dual mode adaptive signal processing technique using different time constants for signal smoothing and system identification is developed. An example of dual mode adaptive signal processing applied to the nonlinear echo cancellation is discussed in detail.

Passive synthetic acoustic arrays for ocean acoustics applications.

This paper gives a brief review, based on the concept of the inverse radiation problem, on the formulation of a synthetic virtual receiving array formed by scanning a single sensor following a designed locus in a stationary acoustic field. Both the beamforming method and the signal decomposition analysis are used to examine the parameters needed for determining the radiation sources’ locations. Studies show that considerable angular resolution can be achieved for observing distant sound sources through adaptive signal processing techniques. Applications in ocean acoustics for directional acoustic field measurement are discussed for two special types of synthetic arrays, linear and circular. Problems involved with the directionality ambiguity due to Doppler frequency shift, angular resolution degradation originated from array deformation, effects caused by instability of ocean conditions, and deployment feasibility are also addressed. The concluding results may be used to establish the guidelines for engineering design and deployment of passive synthetic arrays for ocean acoustics applications.

Atrial electrograms and the characterization of atrial fibrillation.

An improved method of calculating magnitude-squared coherence spectra on pairs of short-duration electrogram recordings is discussed. This method is based on adaptive signal processing techniques and yields spectra with higher resolution than those obtained using a straightforward direct method. The high-resolution spectra will make it possible to examine the time-varying relationship between activity at two sites during atrial fibrillation and may be useful for quick rhythm characterization by implanted devices or for constructing coherence maps in research studies. Example high-resolution spectra for sinus rhythm, atrial flutter, and atrial fibrillation are presented.

Basic study on active noise barrier.

As a trial to improve the performance of passive noise barriers by applying the active control technique, basic theoretical and experimental investigation has been made. At first, the basic conception of the “active noise barrier” is presented, in which an active control system composed of a noise detector, an error sensor, a secondary sound source and an adaptive signal processing unit is attached to a passive barrier. As a preliminary study, a simple numerical calculation was performed according to the Fresnel-Kirchhoff diffraction theory, and the adaptability of the active control has been basically confirmed. Next, 1/2 scale model experiments were performed in a hemi-anechoic room to inve stigatethe effectiveness of the active noise barrier. In this experimental study, the method combining the “filtered x LMS” algorithm and the “block processing technique” was applied as the adaptive signal processing technique. As a result, it has been found that the active noise barrier is effective for sound attenuation in low frequencies.

Active noise control of time‐varying signals in ducts

A sound cancellation system has been developed to reduce the flow‐induced noise in piping systems under both steady‐state and transient conditions. A hybrid controller that combines both open‐loop and closed‐loop responses with adaptive digital signal processing and filtering techniques is utilized to speed up the response of the system controller for the transient condition. Autoregressive, moving average with exogenous input (ARMAX) alogrithm has been utilized in the on‐line and off‐line modeling of signal characteristics. Off‐line model data are introduced into an on‐line adaptive system model via a parameter leakage mechanism. The convergence of the resulting algorithm is improved at the expense of bias in the final system parameters. The bias depends on a weighting or leakage factor, which controls the relative amount of off‐line data to be used. A preliminary test has been conducted to demonstrate the active cancellation concept for synthesized steady‐state signals. Results showed that up to a 30‐dB reduction has been obtained for pure sinusoidal waves ranging from 100 to 500 Hz. For confined jet noise, the noise reduction is about 10 dB, and similar results have been observed for the random broad‐band noise.

Application of recursive least squares algorithm for learning and mathematical reasoning

In our recent publications, we have illustrated how pattern recognition and adaptive signal-processing techniques can be used to construct a machine learning system (MLS). The MLS is based on the manipulation of raw data to extract a set of salient features which have strong significance in the behaviour of the system or device under observation. This paper illustrates the application of the recursive least squares (RLS) algorithm, which has been used as a learning strategy in the MLS, for feature extraction and mathematical reasoning. Here a data-analysis program is developed, which can be used to explain the relationships between the inputs and the outputs of a system.

Adaptive Signal Processing Techniques: An Application to Raman Spectroscopy

In this study, the applicability of adaptive filters and algorithms to the enhancement of Raman spectra is investigated. The two adaptive-filter realizations described are the transversal (tapped-delay) and the lattice. The two adaptive algorithms considered are the least-mean-square and the modified least-square. Experimental results demonstrate that enhancement of Raman spectra obtained at low signal-to-noise ratios can be achieved. Extrapolation is applied to enhance the inherently low resolution of Raman spectra. Adaptive filtering techniques yield results comparable to those obtained from maximum-entropy methods for estimating spectra, with a resolution exceeding that allowed by Fourier procedures.