Log Spectrum Research Articles

Hyperspectral Prediction Models of Chlorophyll Content in Paulownia Leaves under Drought Stress.

This study explored the quantitative inversion of the chlorophyll content in Paulownia seedling leaves under drought stress and analyzed the factors influencing the chlorophyll content from multiple perspectives to obtain the optimal model. Paulownia seedlings were selected as the experimental materials for the potted water control experiments. Four drought stress treatments were set up to obtain four types of Paulownia seedlings: one pair of top leaves (T1), two pairs of leaves (T2), three pairs of leaves (T3), and four pairs of leaves (T4). In total, 23 spectral transformations were selected, and the following four methods were adopted to construct the prediction model, select the best spectral preprocessing method, and explore the influence of water bands: partial least squares modeling with all spectral bands (all-band partial least squares, AB-PLS), principal component analysis partial least squares (PCA-PLS), correlation analysis partial least squares (CA-PLS), correlation analysis (water band) partial least squares, ([CA(W)-PLS]), and vegetation index modeling. Based on the prediction accuracy and the uniformity of different leaf positions, the optimal model was systematically explored. The results of the analysis of spectral reflectance showed significant differences at different leaf positions. The sensitive bands of chlorophyll were located near 550 nm, whereas the sensitive bands of water were located near 1440 and 1920 nm. The results of the vegetation index models indicate that the multiple-index models outperformed the single-index models. Accuracy decreased as the number of indicators decreased. We found that different model construction methods matched different optimal spectral preprocessing methods. First derivative spectra (R') was the best preprocessing method for the AB-PLS, PCA-PLS, and CA-PLS models, whereas the inverse log spectra (log(1/R)) was the best preprocessing method for the CA(W)-PLS model. Among the 14 indices, the green normalized difference vegetation index (GNDVI) was most correlated with the chlorophyll content sensitivity indices, and the water index (WI) was most correlated with the water sensitive indices. At the same time, the water band affected the cross validation accuracy. When characteristic bands were used for modeling, the cross validation accuracy was significantly increased. In contrast, when vegetation indices were used for modeling, the accuracy of the cross validation increased slightly but its predictive ability was reduced; thus, these changes could be ignored. We found that leaf position also affected the prediction accuracy, with the first pair of top leaves exhibiting the worst predictive ability. This was a bottleneck that limited predictive capability. Finally, we found that the CA(W)-PLS model was optimal. The model was based on 23 spectral transformations, four PLS construction methods, water bands, and different leaf positions to ensure systematicity, stability, and applicability.

What is the Simplest Linear Ramp?

We discuss conditions under which a deterministic sequence of real numbers, interpreted as the set of eigenvalues of a Hamiltonian, can exhibit features usually associated to random matrix spectra. A key diagnostic is the spectral form factor (SFF) — a linear ramp in the SFF is often viewed as a signature of random matrix behavior. Based on various explicit examples, we observe conditions for linear and power law ramps to arise in deterministic spectra. We note that a very simple spectrum with a linear ramp is En ~ log n. Despite the presence of ramps, these sequences do not exhibit conventional level repulsion, demonstrating that the lore about their concurrence needs refinement. However, when a small noise correction is added to the spectrum, they lead to clear level repulsion as well as the (linear) ramp. We note some remarkable features of logarithmic spectra, apart from their linear ramps: they are closely related to normal modes of black hole stretched horizons, and their partition function with argument s = β + it is the Riemann zeta function ζ(s). An immediate consequence is that the spectral form factor is simply −ζ|(it)|2. Our observation that log spectra have a linear ramp, is closely related to the Lindelöf hypothesis on the growth of the zeta function. With elementary numerics, we check that the slope of a best fit line through |ζ(it)|2 on a log-log plot is indeed 1, to the fourth decimal. We also note that truncating the Riemann zeta function sum at a finite integer N causes the would-be-eternal ramp to end on a plateau.

Necessity of Log(1/R) and Kubelka-Munk transformation in chemometrics analysis to predict white rice flour adulteration in brown rice flour using visible-near-infrared spectroscopy

This study compared the calibration model performance of reflectance to absorbance transformation spectra combined with pre-processing spectra to find the best model to predict white rice flour adulteration in brown rice flour using the visible and near-infrared spectrometer. Partial least squares regression (PLSR) and principal component regression (PCR) were compared using reflectance, Kubelka-Munk (KM), and Log(1/R) spectra. Area normalization (AN) and Savitsky-smoothing Golay's (SGS) were pre-processing methods. The sample was white rice flour mixed with brown rice flour at 0%, 5%, 10%, 15%, 20%, and 25%. Reflectance spectra outperformed KM and log (1/R) spectra in this study. Reflectance spectra provided the best model for PLSR and PCR. Pre-processed SGS spectra were best for PLSR, while raw reflectance spectra were best for PCR. PLSR and PCR both had an R2 of prediction of 0.96, while the overall average R2 of prediction favors PLSR over PCR. The present study led to the discovery of a simple novel method for developing adulteration flour and showed that a visible near-infrared spectrometer combined with PLSR, or PCR, could predict white rice flour adulteration in brown rice flour.

A new angle-domain cepstral method for generalised gear diagnostics under constant and variable speed operation

This paper proposes a gear diagnostic method denoted as the angle cepstrum comb liftering (ACCL) method. It uses comb liftering in the cepstrum of the order tracked acceleration signal to select all rahmonics of the gearmesh quefrency of a particular meshing gear pair. When converted to the equivalent log spectrum, it thus contains only the harmonics of both gear speeds, and in particular the gearmesh harmonics. The spectrum could thus be adjusted to be periodic at the gearmesh frequency. This implies that the technique not only removes dynamic transfer function effects, independently of speed, but also the (purely static) “mesh transfer functions” (MTFs), defined in Mark’s classic papers on gear vibrations, to give a forcing function very closely related to the combined elemental transmission error (TE) components, both geometric and due to elastic deflections. The paper first develops the method, including the theoretical basis for the removal of the transfer functions, and important practical considerations. The method is then applied to several sets of data corresponding to a range of different faults in gear pairs, including tooth root cracks, and pitting due to surface failure. It was found that Mark’s ALR (average log ratio) method, applied to the cepstrally edited signal, was a natural and appropriate way to evaluate the results, and track fault development. By combining the modified amplitude spectrum with the response phase spectrum, the newly proposed method also allows extraction of angular-domain waveforms similar to the classic TSA (time synchronous average) for each gear, and also the residual signal obtained by removing gearmesh harmonics and one or two pairs of sidebands around them. These are however “closer to the source”, without resonant response effects, and thus not dependent on the operating conditions. The newly proposed approach was found to give similar results for different speeds, and even speed variations up to ±20%, making it a very powerful generalised diagnostic tool. When combined with the original ALR approach, the method is particularly effective for local faults, such as tooth root cracks and presumably large localised spalls, but also for uniformly distributed pitting. In the latter case, it was possible to find parameters that correlate well with indicators of pitting severity, as measured from direct observations of the tooth surface topography. One gear diagnostic technique excluded by this approach is that based on resonant responses to sharp impacts (similar to bearing fault diagnosis), which has been found in some cases to give earlier warning of incipient cracks, for which a different approach would be required. Another is uniform abrasive wear of profiles, but this can often be easily diagnosed by changes in gearmesh harmonics and close sidebands, for which the proposed approach might also be found useful.

AdaptSPEC-X: Covariate-Dependent Spectral Modeling of Multiple Nonstationary Time Series

We present the AdaptSPEC-X method for the joint analysis of a panel of possibly nonstationary time series. The approach is Bayesian and uses a covariate-dependent infinite mixture model to incorporate multiple time series, with mixture components parameterized by a time-varying mean and log spectrum. The mixture components are based on AdaptSPEC, a nonparametric model which adaptively divides the time series into an unknown number of segments and estimates the local log spectra by smoothing splines. AdaptSPEC-X extends AdaptSPEC in three ways. First, through the infinite mixture, it applies to multiple time series linked by covariates. Second, it can handle missing values, a common feature of time series which can cause difficulties for nonparametric spectral methods. Third, it allows for a time-varying mean. Through these extensions, AdaptSPEC-X can estimate time-varying means and spectra at observed and unobserved covariate values, allowing for predictive inference. Estimation is performed by Markov chain Monte Carlo (MCMC) methods, combining data augmentation, reversible jump, and Riemann manifold Hamiltonian Monte Carlo techniques. We evaluate the methodology using simulated data, and describe applications to Australian rainfall data and measles incidence in the United States. Software implementing the method proposed in this article is available in the R package BayesSpec. Supplementary files for this article are available online.

Electronic, electrical and dielectric analysis of Cr-doped hydroxyapatite

The electrical and the dielectric response of synthesized hydroxyapatite (HA) and Cr doped HA samples are obtained by impedance spectroscopy analysis. Variation in conductivity of materials was associated with the mutual interaction of Cr(III) and Cr(IV) with OH vacancies. The observed έ_LF and έ_HF dispersions of log έ vs log(f) spectra of HA and Cr-HA-100 are related to dipolar relaxations of H+ and OH− ions. The interfacial polarization is responsible for the colossal magnitude of dielectric constant and dielectric loss. This study suggests that the controlled ionic conduction, microstructure, and interfacial effects could produce excellent electrical and dielectric properties.

Development of NIR-HSI and chemometrics process analytical technology for drying of beef jerky

Beef jerky samples immersed in brine or water, both with and without ultrasound treatment were dried at 60 °C for 30, 60, 90, 120, 150, 180, 210 and 240 min. Drying behaviour was evaluated using ten drying kinetic models and hyperspectral imaging. All samples reached water activity values (Aw) < 0.85 during the first 90 min of drying. Moisture content (MC) of beef jerky was predicted using near infrared hyperspectral imaging and chemometrics. Partial least squares regression, band selection and spectral pre-treatments were applied to develop MC prediction models using beef jerky spectra. Most of the MC prediction models developed in this work had RPD values >4, indicating their suitability for process control applications. The best performing MC prediction models for the non-ultrasound and ultrasound treated samples were developed using the ensemble Monte Carlo variable selection (EMCVS) on second derivative of log(1/R) spectra and EMCVS-selectivity ratio on linear detrended log (1/R) spectra, respectively. This study demonstrated the potential of NIR-HSI and chemometrics as a PAT tool for drying of beef jerky.

Application of Neural Network Algorithm Based on Principal Component Image Analysis in Band Expansion of College English Listening.

With the development of information technology, band expansion technology is gradually applied to college English listening teaching. This technology aims to recover broadband speech signals from narrowband speech signals with a limited frequency band. However, due to the limitations of current voice equipment and channel conditions, the existing voice band expansion technology often ignores the high-frequency and low-frequency correlation of the audio, resulting in excessive smoothing of the recovered high-frequency spectrum, too dull subjective hearing, and insufficient expression ability. In order to solve this problem, a neural network model PCA-NN (principal components analysis-neural network) based on principal component image analysis is proposed. Based on the nonlinear characteristics of the audio image signal, the model reduces the dimension of high-dimensional data and realizes the effective recovery of the high-frequency detailed spectrum of audio signal in phase space. The results show that the PCA-NN, i.e., neural network based on principal component analysis, is superior to other audio expansion algorithms in subjective and objective evaluation; in log spectrum distortion evaluation, PCA-NN algorithm obtains smaller LSD. Compared with EHBE, Le, and La, the average LSD decreased by 2.286 dB, 0.51 dB, and 0.15 dB, respectively. The above results show that in the image frequency band expansion of college English listening, the neural network algorithm based on principal component analysis (PCA-NN) can obtain better high-frequency reconstruction accuracy and effectively improve the audio quality.

On accumulation points of pseudo-effective thresholds

We characterize a $k$-th accumulation point of pseudo-effective thresholds of $n$-dimensional varieties as certain invariant associates to a numerically trivial pair of an $(n-k)$-dimensional variety. This characterization is applied towards Fujita's log spectrum conjecture for large $k$.

Diffusion vs. fluid alteration in alkali feldspar 40Ar/39Ar thermochronology: Does cross-correlation of log(r/r0) and age spectra validate thermal histories?

For six decades geoscientists have been trying to quantitatively understand the nature of radiogenic Ar loss from alkali feldspar. Some researchers suggest that volume diffusion is the dominant mechanism, and they use conventional step-heating 40Ar/39Ar data from alkali feldspar to recover the thermal histories of rocks. They argue that a high degree of correlation between log(r/r0) and 40Ar/39Ar age spectra, which is observed in a number of natural examples, justifies this hypothesis. In contrast, other investigators suggest that fluid-mediated recrystallisation and alteration control the radiogenic Ar redistribution, hence rendering alkali feldspar useless as a thermochronometer. By means of numerical modelling, we found that the latter mechanism as well is able to produce samples with highly correlated log(r/r0) and 40Ar/39Ar age spectra. In addition, we show that apparent thermal histories recovered for altered alkali feldspar crystals by interpreting step-heating 40Ar/39Ar data may be grossly inaccurate, and yet seemingly fit the prevailing understanding of regional geology. Such inaccurate apparent thermal histories can be obtained even from alkali feldspar crystals that underwent volumetrically low degrees of alteration. Therefore, we conclude that conventional step-heating 40Ar/39Ar data are insufficient to support the assumption that radiogenic Ar loss from alkali feldspar occurred solely by volume diffusion and validate the constrained thermal histories, even if upheld by a priori knowledge of regional tectonics. We further suggest that all thermochronological constraints obtained using such data should be supported by detailed petrological characterisation of alkali feldspar.

On Fujita’s conjecture for pseudo-effective thresholds

We show Fujita's spectrum conjecture for $\epsilon$-log canonical pairs and Fujita's log spectrum conjecture for log canonical pairs. Then, we generalize the pseudo-effective threshold of a single divisor to multiple divisors and establish the analogous finiteness and the DCC properties.

Convolutional neural network for single-sensor acoustic localization of a transiting broadband source in very shallow water

When a broadband source of radiated noise transits past a fixed hydrophone, a Lloyd's mirror constructive/destructive interference pattern can be observed in the output spectrogram. By taking the spectrum of a (log) spectrum, the power cepstrum detects the periodic structure of the Lloyd's mirror fringe pattern by generating a sequence of pulses located at the fundamental quefrency and its multiples. The fundamental quefrency, which is the reciprocal of the frequency difference between adjacent destructive interference fringes, equates to the multipath delay time. An experiment is conducted where a motorboat transits past a hydrophone located about 1 m above the seafloor in very shallow water (20 m deep). The hydrophone has a frequency bandwidth of 90 kHz, and its output is sampled at 250 kHz. A cepstrogram database is compiled from multiple vessel transits, and its cepstrum-based feature vectors (along with ground-truth range data) form the input to train a convolutional neural network (CNN) so that it can predict the source range relative to the hydrophone for other ("unseen") vessel transits. The CNN provides an accurate prediction of the instantaneous source range even when the range estimate from conventional multipath passive ranging is biased, which occurs at low grazing angles (<12°).

Hydrodynamic and Hydrographic Modeling of Istanbul Strait

The aim of this study is to model the hydrodynamic processes of the Istanbul Strait with its stratified flow characteristics, and calibrate the most important parameters using local and global search algorithms. For that, two open boundary conditions are defined, which are in the northern and southern parts of the Strait. Observed bathymetric, hydrographic, meteorological, and water-level data are used to set up the Delft3D-FLOW model. First, the sensitivities of the model parameters on the numerical model outputs are assessed using Parameter EStimation Tool (PEST) toolbox. Then, the model is calibrated based on the objective functions, focusing on the flow rates of the upper and lower layers. The salinity and temperature profiles of the strait are only used for model validation. The results show that the calibrated model outputs of the Istanbul Strait are reliable and consistent with the in situ measurements. The sensitivity analysis reveals that the spatial low-pass filter coefficient, horizontal eddy viscosity, Prandtl–Schmidt number, slope in log–log spectrum, and Manning roughness coefficient are most sensitive parameters affecting the flow rate performance of the model. The agreement between observed salinity profiles and simulated model outputs is promising, whereas the match between observed and simulated temperature profiles is weak, showing that the model can be improved, particularly for simulating the mixing layer.

Modulation-Domain Kalman Filtering for Monaural Blind Speech Denoising and Dereverberation

We describe a monaural speech enhancement algorithm based on modulation-domain Kalman filtering to blindly track the time–frequency log-magnitude spectra of speech and reverberation. We propose an adaptive algorithm that performs blind joint denoising and dereverberation, while accounting for the inter-frame speech dynamics, by estimating the posterior distribution of the speech log-magnitude spectrum given the log-magnitude spectrum of the noisy reverberant speech. The Kalman filter update step models the non-linear relations between the speech, noise, and reverberation log spectra. The Kalman filtering algorithm uses a signal model that takes into account the reverberation parameters of the reverberation time $T_{60}$ and the direct-to-reverberant energy ratio (DRR) and also estimates and tracks $T_{60}$ and the DRR in every frequency bin to improve the estimation of the speech log spectrum. The proposed algorithm is evaluated in terms of speech quality, speech intelligibility, and dereverberation performance for a range of reverberation parameters and reverberant speech to noise ratios, in different noises, and is also compared to competing denoising and dereverberation techniques. Experimental results using noisy reverberant speech demonstrate the effectiveness of the enhancement algorithm.

The bias of the log power spectrum for discrete surveys

Abstract A primary goal of galaxy surveys is to tighten constraints on cosmological parameters, and the power spectrum P(k) is the standard means of doing so. However, at translinear scales P(k) is blind to much of these surveys’ information – information which the log density power spectrum recovers. For discrete fields (such as the galaxy density), A* denotes the statistic analogous to the log density: A* is a ‘sufficient statistic’ in that its power spectrum (and mean) capture virtually all of a discrete survey's information. However, the power spectrum of A* is biased with respect to the corresponding log spectrum for continuous fields, and to use $P_{A^*}(k)$ to constrain the values of cosmological parameters, we require some means of predicting this bias. Here, we present a prescription for doing so; for Euclid-like surveys (with cubical cells 16h−1 Mpc across) our bias prescription's error is less than 3 per cent. This prediction will facilitate optimal utilization of the information in future galaxy surveys.

Optimal Bandwidth for Multitaper Spectrum Estimation

A systematic method for bandwidth parameter selection is desired for Thomson multitaper spectrum estimation. We give a method for determining the optimal bandwidth based on a mean squared error (MSE) criterion. When the true spectrum has a second-order Taylor series expansion, one can express quadratic local bias as a function of the curvature of the spectrum, that can be estimated by using a simple spline approximation. This is combined with a variance estimate, obtained by jackknifing over individual spectrum estimates, to produce an estimated MSE for the log spectrum estimate for each choice of time-bandwidth product. The bandwidth that minimizes the estimated MSE then gives the desired spectrum estimate. Additionally, the bandwidth obtained by using our method is also optimal for cepstrum estimates. We give an example of a damped oscillatory (Lorentzian) process in which the approximate optimal bandwidth can be written as a function of the damping parameter. The true optimal bandwidth agrees well with that given by minimizing estimated the MSE in these examples.

Simple but efficient signal pre-processing in soil organic carbon spectroscopic estimation

As there is no single (or combination of) signal pre-processing method that works best with all data sets, choosing the most feasible one is a key aspect in soil diffuse reflectance spectroscopy in the visible– and near infrared region (400–2500nm). The commonly used pre-processing methods include tools for spectra smoothing and/or noise reduction (e.g. Savitzky-Golay (SG) filtering or discrete wavelet transformation (DWT)), light scatter correction (multiplicative scatter correction (MSC), standard normal variate (SNV)), baseline normalization techniques to cope with vertical offset and/or slope effects (e.g. continuum removal (CR), first and second order derivative (FD and SD)), as well as other transformations (e.g. logarithmic-log(1/R)). All of these tools are aimed at eliminating or reducing unwanted side effects (artifacts) in the spectra and at enhancing the recognition of relevant information. For soil organic carbon content estimation using partial least square regression calibration technique, smoothing with SG filter and (or in combination with) CR usually ensures a reliable estimation. However, the common CR may suffer from a few shortcomings. An approximation is applied to connect the pivot points of the spectrum in order to derive a continuum, but more problematically, the CR procedure does not recognize the true essence of the vertical shift at the very beginning of the spectra (the CR value always equals one at that point). Therefore, we decided to modify the procedure in the way that the reflectance values at respective wavelengths were divided not by the continuum, but by the maximal reflectance value of the particular spectrum. This correction by the maximum reflectance (CMR) pre-processing was tested in comparison with eight other above mentioned methods at four different study sites that differ in the prevailing soil units. As a result, on site 1 (Haplic Chernozem), we achieved a significantly improved prediction accuracy using the CMR (R2cv=0.845) compared to raw (but smoothed) soil spectra (0.815). On site 2 (Rendzic Leptosol), the most accurate prediction was achieved equally with CMR, MSC, SNV, log(1/R), DWT and raw spectra (R2cv from 0.560 to 0.592), and on site 3 (Haplic Cambisol) equally with MSC and CMR (both R2cv=0.767), as only these two were significantly different from the raw spectra. On site 4 (Haplic Luvisol), the only one significantly more accurate prediction compared to raw spectra was achieved with FD (R2cv=0.611), while for the rest of the methods, except SD, there was no difference if either raw spectra or other transformations were used (R2cv from 0.499 to 0.591). Finally, using the whole data set the differences between pre-processing methods were even less pronounced, when there was no significant difference between raw spectra and other methods (except SD which was significantly worse), although all the predictions were more accurate in general (R2cv from 0.811 to 0.831).

A history of cepstrum analysis and its application to mechanical problems

It is not widely realised that the first paper on cepstrum analysis was published two years before the FFT algorithm, despite having Tukey as a common author, and its definition was such that it was not reversible even to the log spectrum. After publication of the FFT in 1965, the cepstrum was redefined so as to be reversible to the log spectrum, and shortly afterwards Oppenheim and Schafer defined the “complex cepstrum”, which was reversible to the time domain. They also derived the analytical form of the complex cepstrum of a transfer function in terms of its poles and zeros. The cepstrum had been used in speech analysis for determining voice pitch (by accurately measuring the harmonic spacing), but also for separating the formants (transfer function of the vocal tract) from voiced and unvoiced sources, and this led quite early to similar applications in mechanics. The first was to gear diagnostics (Randall), where the cepstrum greatly simplified the interpretation of the sideband families associated with local faults in gears, and the second was to extraction of diesel engine cylinder pressure signals from acoustic response measurements (Lyon and Ordubadi). Later Polydoros defined the differential cepstrum, which had an analytical form similar to the impulse response function, and Gao and Randall used this and the complex cepstrum in the application of cepstrum analysis to modal analysis of mechanical structures. Antoni proposed the mean differential cepstrum, which gave a smoothed result. The cepstrum can be applied to MIMO systems if at least one SIMO response can be separated, and a number of blind source separation techniques have been proposed for this. Most recently it has been shown that even though it is not possible to apply the complex cepstrum to stationary signals, it is possible to use the real cepstrum to edit their (log) amplitude spectrum, and combine this with the original phase to obtain edited time signals. This has already been used for a wide range of mechanical applications. A very powerful processing tool is an exponential “lifter” (window) applied to the cepstrum, which is shown to extract the modal part of the response (with a small extra damping of each mode corresponding to the window). This can then be used to repress or enhance the modal information in the response according to the application.

Dithering techniques in automatic recognition of speech corrupted by MP3 compression: Analysis, solutions and experiments

A large portion of the audio files distributed over the Internet or those stored in personal and corporate media archives are in a compressed form. There exist several compression techniques and algorithms but it is the MPEG Layer-3 (known as MP3) that has achieved a really wide popularity in general audio coding, and in speech, too. However, the algorithm is lossy in nature and introduces distortion into spectral and temporal characteristics of a signal. In this paper we study its impact on automatic speech recognition (ASR). We show that with decreasing MP3 bitrates the major source of ASR performance degradation is deep spectral valleys (i.e. bins with almost zero energy) caused by the masking effect of the MP3 algorithm. We demonstrate that these unnatural gaps in spectrum can be effectively compensated by adding a certain amount of noise to the distorted signal. We provide theoretical background for this approach where we show that the added noise affects mainly the spectral valleys. They are filled by the noise while the spectral bins with speech remain almost unchanged. This helps to restore a more natural shape of log spectrum and cepstrum, and consequently has a positive impact on ASR performance. In our previous work, we have proposed two types of the signal dithering (noise addition) technique, one applied globally, the other in a more selective way. In this paper, we offer a more detailed insight into their performance. We provide results from many experiments where we test them in various scenarios, using a large vocabulary continuous speech recognition (LVCSR) system, acoustic models based on gaussian-mixture model (GMM) as well as on deep-neural network (DNN), and multiple speech databases in three languages (Czech, English and German). Our results prove that both the proposed techniques, and the selective dithering method, in particular, yield consistent compensation of the negative impact of the MP3 compressed speech on ASR performance.

Foreground Speech Segmentation and Enhancement Using Glottal Closure Instants and Mel Cepstral Coefficients

In this paper, the speech signal recorded from the desired speaker close to microphone in natural environment is regarded as foreground speech and rest of the interfering sources as background noise . The proposed paper exploits speech production features like glottal closure instants in time domain and vocal tract information in spectral domain to segment the desired speaker's speech and to further enhance it. The foreground speech is perceptually enhanced using the auditory perception feature in mel-frequency domain using mel-cepstral coefficients and its inversion using mel log spectrum approximation filter. The focus is on enhancing the production and perceptual features of foreground speech rather than relying on modeling the interfering sources. The speech data are collected in different natural environments from different speakers in order to evaluate the proposed method. The enhanced speech signals derived at three different stages of the proposed method are evaluated with state-of-the-art methods in terms of subjective and objective measures. The proposed method provides improved performance compared to the considered state-of-the-art methods. In terms of the proposed objective measure foreground to background Ratio , the enhancement approach presented in this paper gives an average improvement of 12 dB as opposed to existing spectral subtraction-based method which provides 3 dB. Moreover, subjective evaluation using 24 different subjects corroborates the objective test results.