Noise Spectral Estimation Research Articles

Spectral precision of frequency demodulation method: Influence of additive noise on instantaneous angular speed spectral estimation

This paper treats the spectral precision of frequency demodulation method based on Gabor’s analytic signal applied to Instantaneous Angular Speed (IAS) estimation from speed sensor under stationary speed conditions. The quadrature decomposition combined with geometric complex representation are used to approach the spectral modeling of the influence of wide-sense stationary additive noise on IAS spectral estimate using the periodogram. A method is proposed for IAS noise spectral estimation from signal’s envelope. Moreover, the effect of the carrier power and encoder resolution is demonstrated through real life signals. Finally, a statistical approach is proposed in order to set thresholds on IAS spectral observation allowing a good separation of additive random perturbations from real information under a certain probability. The proposed tools are applied on both synthetic and experimental signals and are shown to be effective in estimating IAS noise level from the input signal properties.

Noise spectral analysis and error estimation of continuous glucose monitors under real-life conditions of diabetes patients

We present an analysis of the nature of noise and measurement errors for two popular continuous glucose monitors (Medtronic OneTouch UltraSmart and Abbott FreeStyle Libre) under daily, real-life conditions. We use three different techniques to extract the noise in the time series of the interstitial glucose values provided by these sensors: local regression, singular spectrum analysis, and wavelets. Likewise, we characterize the noise using five different methods: power spectral density, normality tests for noise values, the generalized Hurst exponent, Bayesian information criterion, and the fit of the noise to an autoregressive model. To analyze measurement errors, we use five different metrics: the instantaneous error, the minimum error in a 60-min timeframe, the lag of the minimum error in the 60-min timeframe, and both the average and the standard deviation of the absolute value of the instantaneous error. We perform the analysis on data provided by 20 patients totaling 198162 samples that account for more than 1732 days of data measured by patients during their daily lives. The results conclude that continuous glucose monitors have either pink or red noise; the fit of autoregressive models to the noise is lower than previously reported, and the elective model, if any, has a high order; the medians of the measurement errors have a strong presence of outliers with values above 80 mg/dl, lags higher than reported in hospital, controlled trials, and considerable intra- and inter-patient variability; and that glucose monitor accuracy is dependent upon the patients’ glycemic variability.

Noise spectral estimation methods and their impact on gravitational wave measurement of compact binary mergers

Estimating the parameters of gravitational wave signals detected by ground-based detectors requires an understanding of the properties of the detectors' noise. In particular, the most commonly used likelihood function for gravitational wave data analysis assumes that the noise is Gaussian, stationary, and of known frequency-dependent variance. The variance of the colored Gaussian noise is used as a whitening filter on the data before computation of the likelihood function. In practice the noise variance is not known and it evolves over timescales of dozens of seconds to minutes. We study two methods for estimating this whitening filter for ground-based gravitational wave detectors with the goal of performing parameter estimation studies. The first method uses large amounts of data separated from the specific segment we wish to analyze and computes the power spectral density of the noise through the mean-median Welch method. The second method uses the same data segment as the parameter estimation analysis, which potentially includes a gravitational wave signal, and obtains the whitening filter through a fit of the power spectrum of the data in terms of a sum of splines and Lorentzians. We compare these two methods and argue that the latter is more reliable for gravitational wave parameter estimation.

A Speech Denoising Algorithm Based on Harmonic Regeneration

Single channel speech enhancement in noisy environments is studied in this paper. Traditional speech enhancement methods that based on short time spectral analysis frequently introduce harmonic distortion due to imprecise noise spectral estimation. A kind of technology which is called harmonic regeneration is used to overcome the defect of traditional speech enhancement methods, and damaged harmonics are restored. Firstly, a non-linear operation is used on the harmonic distorted signal to generate an artificial signal which contains all of harmonics of pure speech signal. Secondly, the artificial signal is used to correct the traditional noise reduction gain function, and the purpose of harmonic reservation is reached. Theoretical analysis, objective and subjective results indicate that the harmonic regeneration algorithm can improve speech quality dramatically. The computational complexity of the new algorithm is rather low, and it can be easily implemented in real time applications.

A novel fast nonstationary noise tracking approach based on MMSE spectral power estimator

Estimating the noise power spectral density (PSD) from the corrupted speech signal is an essential component for speech enhancement algorithms. In this paper, a novel noise PSD estimation algorithm based on minimum mean-square error (MMSE) is proposed. The noise PSD estimate is obtained by recursively smoothing the MMSE estimation of the current noise spectral power. For the noise spectral power estimation, a spectral weighting function is derived, which depends on the a priori signal-to-noise ratio (SNR). Since the speech spectral power is highly important for the a priori SNR estimate, this paper proposes an MMSE spectral power estimator incorporating speech presence uncertainty (SPU) for speech spectral power estimate to improve the a priori SNR estimate. Moreover, a bias correction factor is derived for speech spectral power estimation bias. Then, the estimated speech spectral power is used in “decision-directed” (DD) estimator of the a priori SNR to achieve fast noise tracking. Compared to three state-of-the-art approaches, i.e., minimum statistics (MS), MMSE-based approach, and speech presence probability (SPP)-based approach, it is clear from experimental results that the proposed algorithm exhibits more excellent noise tracking capability under various nonstationary noise environments and SNR conditions. When employed in a speech enhancement system, improved speech enhancement performances in terms of segmental SNR improvements (SSNR+) and perceptual evaluation of speech quality (PESQ) can be observed.

The spectral profile estimation algorithm: a non-linear, non-a priori noise normalization algorithm

Time-frequency analysis of acoustic signals often involves a background noise spectral estimation step to estimate the signal-to-noise ratio of the signals of interest. This step is typically accomplished with the application of a split mean normalizer (SPN) [Struzinski and Lowe, J. Acoust. Soc. Am. 76 1738–1742 (1984)]. These normalizers work well against tones and can be tuned with a priori information to normalize wider energy signals but can still suffer performance loss in these cases. For signal dense spectra that contain both tonal and broadband-like energy the SPN has difficulty separating noise from signal, and the signal energy unduly influences the background estimate. This work introduces a new normalization scheme called the Spectral Profile Estimation (SPE) algorithm, which operates with no a priori information to estimate the background noise of these signal dense spectra. The SPE algorithm assumes that the background noise has an f-m, m>0, shape, and that signals are superimposed on the noise and are associated with spectra maxima. The SPE algorithm then finds and connects local minima to estimate the background. The SPE algorithm is first explained and then applied to experimental data. Finally, SPE performance is compared to the performance of SPN.

HMM-Based Reconstruction of Unreliable Spectrographic Data for Noise Robust Speech Recognition

This paper presents a framework for efficient HMM-based estimation of unreliable spectrographic speech data. It discusses the role of hidden Markov models (HMMs) during minimum mean-square error (MMSE) spectral reconstruction. We develop novel HMM-based reconstruction algorithms which exploit intra-channel (across-time) correlation and/or inter-channel (across-frequency) correlation. For the sake of computational efficiency, this paper utilizes approximations to HMM-based decoding methods by developing models constructed from lower resolution quantizers. State configurations for lower resolution models are obtained through a tree-structured mapping of quantizer centroids, and model parameters are adapted accordingly. HMM downsampling avoids expensive retraining of models, and eliminates unnecessary memory requirements. Explicit general formulae are presented for the adaptation of steady-state and transitional statistics. Adaptation of observation statistics are derived from stochastic models of noise spectral magnitude estimation accuracies. The proposed estimation methods are applied in combination with oracle masks, which provide an upper performance bound, as well as masks derived from speech presence probability, which represent a more realistic scenario. Both methods are shown to boost noise robust recognition accuracies significantly relative to the Mel-frequency cepstral coefficient (MFCC) baseline system. Furthermore, HMM downsampling greatly reduces the complexity of the HMM-based reconstruction method while negligibly affecting results.

Adaptive noise spectral estimation for spectral subtraction speech enhancement

An adaptive scheme for noise spectral estimation is proposed to cope with the power spectral subtraction method for speech enhancement. The use of such a noise spectral estimate helps eliminate residual noise without increasing speech distortion. Improvements in the segmental signal-to-noise ratio and modified bark spectral distortion measures are evidently seen with 192 TIMIT sentences corrupted by four types of broadband noise taken from the Noisex-92 database. Moreover, incorporation of this noise adaptation scheme into the constrained parameter estimator is demonstrated. Mean opinion score listening tests confirm that the proposed subtractive scheme yields better results than that acquired by subtracting an averaged noise power spectrum.

Optimal processing and performance evaluation of passive acoustic systems

The acoustic spectrogram is a fundamental input for many underwater acoustic signal detectors. This presentation describes an effort to place theoretical bounds on the performance of these detectors and to provide insight into the structure of practical processors. The presentation begins with a description of the Neyman–Pearson optimal detector, the likelihood ratio test, and its use for signal detection in colored Gaussian noise scenarios. The presentation continues by developing successively more complex detection scenarios that necessitate the use of background estimation and, consequently, the generalized likelihood ratio test (GLRT). Parameter estimation from several common noise spectral estimation (NSE) techniques is included in the GLRT detection scenarios. In addition, the presentation develops guidelines for matching NSE parameter settings to signal characteristics. Then, the relationships among several relevant signal-to-noise ratios (SNRs) is developed and their correlation with detection performance is considered. Finally, the presentation considers departures from the Gaussian distribution assumption for background and signal processes and the resulting consequences. a)Ensign, USN

The realization and performance analysis of noise spectral density estimation using a digital bandpass filter

The continuous spectral estimation of non-ARMA stochastic processes (enclosed noise processes), has received relatively little attention in the traditional signal processing literature. The concept of the application of bandpass filters to spectral analysis was introduced more than 25 years ago; however, a practical measurement instrument using bandpass filters can only make noise measurements at a few discrete frequency points. In this paper, a noise spectral density measurement over a wide frequency range, using a recursive digital bandpass filtering algorithm, has been realized. The variance and bias in the equi-bandwidth measurement and the equi-Q; measurement have been discussed. Finally, the algorithm's parameters are decided according to the measurement accuracy and fidelity.