Single-Channel Speech Enhancement

description27 papers

group33 followers

lightbulbAbout this topic

Single-Channel Speech Enhancement is a field of signal processing focused on improving the quality and intelligibility of speech signals captured from a single audio channel. It employs various algorithms to reduce background noise, reverberation, and other distortions, thereby enhancing the clarity of the speech for better communication and recognition.

lightbulbAbout this topic

Key research themes

1. How can deep learning architectures improve estimation of model parameters for enhanced single-channel speech enhancement?

This theme investigates the integration of deep learning with traditional parametric and filtering models (e.g., autoregressive models, Kalman filters) to enhance estimation of speech and noise characteristics in single-channel speech enhancement. This research direction is crucial as accurate estimation of parameters such as linear prediction coefficients (LPCs) directly impacts the quality and intelligibility of enhanced speech, while overcoming limitations of classical methods in noisy, non-stationary environments.

DeepLPC: A Deep Learning Approach to Augmented Kalman Filter-Based Single-Channel Speech Enhancement

by Dr. Sujan K U M A R Roy

2022

Key finding: This work proposes DeepLPC, a deep learning framework that jointly estimates clean speech and noise linear prediction coefficient (LPC) power spectra without relying on whitening filters, thereby significantly reducing bias... Read more

articleView Paper downloadDownload

Integrating Statistical Uncertainty into Neural Network-Based Speech Enhancement

by Stefan Wermter

2024, ICASSP 2022 - 2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

Key finding: This paper introduces a neural network trained to estimate both the Wiener filter and its associated variance (uncertainty) for spectral coefficients in speech enhancement. The approach models the full posterior distribution... Read more

articleView Paper downloadDownload

Multi-Attention Bottleneck for Gated Convolutional Encoder-Decoder-Based Speech Enhancement

by AYMEN TRIGUI

2025, IEEE Access

Key finding: The paper proposes a Multi-Attention Bottleneck (MAB) that integrates a Transformer-based self-attention combined with time-frequency and channel attention modules within a gated convolutional encoder-decoder (CED)... Read more

articleView Paper downloadDownload

Supervised Single Channel Speech Enhancement Method Using UNET

by SHAKHAWAT HOSEN

2023, Electronics

Key finding: This study applies a supervised deep learning approach based on U-Net architectures to single-channel speech enhancement, using magnitude spectrogram inputs and leveraging convolutional and recurrent layers to model both... Read more

articleView Paper downloadDownload

keyboard_arrow_downShow more

2. What signal representations and masking strategies optimize non-negative matrix factorization (NMF)-based single-channel speech enhancement?

This research area focuses on leveraging specialized signal representations (e.g., wavelet transforms) and jointly learning ratio masking functions with dictionary learning within the NMF framework to improve single-channel speech enhancement. Since traditional STFT-based methods face limitations like time-frequency resolution trade-offs and noisy phase estimation, exploring alternative transforms and mask formulations can yield enhanced noise suppression and better preservation of speech components, which is essential for effective and efficient enhancement.

Supervised single-channel speech enhancement using ratio mask with joint dictionary learning

by Shohidul Islam

2022, Speech Communication

Key finding: This work introduces a novel speech enhancement approach applying Dual-Tree Complex Wavelet Transform (DTCWT) for shift-invariant subband decomposition, combined with joint dictionary learning of subband smooth ratio masks... Read more

articleView Paper downloadDownload

Learning speech features in the presence of noise: Sparse convolutive robust non-negative matrix factorization

by Scott Rickard

2015, 2009 16th International Conference on Digital Signal Processing

Key finding: The paper proposes Sparse Convolutive Robust Non-negative Matrix Factorization (SCRNMF), an NMF extension that explicitly models non-stationary noise as an interfering source and learns speech features with temporal extent... Read more

articleView Paper downloadDownload

3. How do advanced signal-domain transformations and statistical models contribute to improved MMSE estimators in single-channel speech enhancement?

Under this theme, research investigates the impact of adopting alternative signal transforms, like the Discrete Cosine Transform (DCT), and statistical speech priors (e.g., Gaussian, Laplacian, Gamma) to derive closed-form Minimum Mean Square Error (MMSE) estimators for short-time spectral amplitude. By overcoming analytical challenges associated with traditional Discrete Fourier Transform (DFT)-based methods and super-Gaussian priors, these studies aim to optimize noise suppression and speech fidelity, which are critical for both objective and subjective speech enhancement outcomes.

On DCT-based MMSE estimation of short time spectral amplitude for single-channel speech enhancement

by Sisi Shi

2022, Elsevier

Key finding: This paper derives closed-form MMSE estimators of clean short-time spectral amplitude in the Discrete Cosine Transform (DCT) domain assuming various speech prior distributions including Gaussian, Laplace, and Gamma, under an... Read more

articleView Paper downloadDownload

All papers in Single-Channel Speech Enhancement

Synergy of acoustic-phonetics and auditory modeling towards robust speech recognition

by Carol Espy-Wilson

2025

The problem addressed in this work is that of enhancing speech signals corrupted by additive noise and improving the performance of automatic speech recognizers in noisy conditions. The enhanced speech signals can also improve the... more

descriptionView Paper arrow_downwardDownload

Phase Estimation in Single Channel Speech Enhancement Using Phase Decomposition

by Andrew Davydov

2024, IEEE Signal Processing Letters

descriptionView Paper arrow_downwardDownload

Separation of human and animal seismic signatures using non-negative matrix factorization

by Asif Mehmood

2023, Pattern Recognition Letters

Seismic footstep detection based systems can be employed for homeland security applications such as perimeter protection and the border security. This paper reports an approach based on non-negative matrix factorization (NMF) for seismic... more

descriptionView Paper arrow_downwardDownload

Performance Analysis of Statistical Approaches and NMF Approaches for Speech Enhancement

by Dr. Ravi Kumar Kandagatla

2023, International Journal of Image, Graphics and Signal Processing

Super-Gaussian Based Bayesian Estimators plays significant role in noise reduction. However, the traditional Bayesian Estimators process only DFT spectral amplitude of noisy speech and the phase is left unprocessed. While deriving... more

descriptionView Paper arrow_downwardDownload

Separation of human and animal seismic signatures using non-negative matrix factorization

by Thyagaraju Damarla

2023, Pattern Recognition Letters

descriptionView Paper arrow_downwardDownload

Embedded CPS for Real-Time Monitoring of a Laser Beam Deflection System Using Spectral Analysis

by JAVIER ALBERTO MARIÑO DIAZ

2023

Cyber-Physical Systems (CPS) are seen as true technology enablers for new complex industrial applications from the perspective of the Industry 4.0. In this context, the challenges of advanced laser material processing applications present... more

descriptionView Paper arrow_downwardDownload

Securing Speech in GSM Networks using DES with Random Permutation and Inversion Algorithm

by Khaled Merit

2023, International Journal of Distributed and Parallel systems

Global System for Mobile Communications (GSM) is one of the most commonly used cellular technologies in the world. One of the objectives in mobile communication systems is the security of the exchanged data. GSM employs many cryptographic... more

descriptionView Paper arrow_downwardDownload

Covert Visual Spatial Attention: Effects Of Voluntary And Involuntary Attention On Channel Enhancement And Channel Selection

by ROCIO LUNA

2022

The ability to fixate ones eyes on one object while attending to another object is known as covert visual attention. The present study investigated the effects of covert visual attention on reaction time (RT) and accuracy while... more

descriptionView Paper arrow_downwardDownload

On DCT-based MMSE estimation of short time spectral amplitude for single-channel speech enhancement

by Sisi Shi

2022, Elsevier

This paper proposes Discrete Cosine Transform (DCT) based speech enhancement algorithms. These algorithms utilize minimum mean square error (MMSE) estimator of clean short-time spectral amplitude, which respectively uses Gaussian, Laplace... more

descriptionView Paper arrow_downwardDownload

On DCT-based MMSE estimation of short time spectral amplitude for single-channel speech enhancement

by Sisi Shi

2022, Elsevier

Fig. 1. Perceived quality estimation for Polarity-Only (PO, solid line), Phase-Only (PhO, dashed line) and noisy (dotted line) stimuli as a function Segmental SNR.

Fig. 18. The total probability of error as a function of Sun. The optimal &,, is found by minimizing the total probability of error when ¢ is uniformly distributed between €,., = 0 and €,, = 100,

Fig. 15. The mean subjective preference score (%) comparison for each speech enhancement method. The female utterance (FBO7_09) corrupted with 5 dB voice babble noise was used for the subjective tests. The error bars indicated the standard deviation of the scores.

Fig. 16. Performance comparison between the state of the art phase-aware estimators (i.e., ADP [51], AUP [27], and MAUP [28]) and the propose estimators [i.e., Gy (15), G, (22), and G¢ (27)] using blind a prior SNR and noise PSD estimation. The clean speech and noisy (unprocessed) speech are also included as the upper bound and lower bound of the performance, respectively. Results are shown in terms of conventional (i.e., PESQ, STOI and Segmental SNR) and phase-aware (i.e., PD, UnRMSE, and UnHPSNR) instrumental metrics. Scores of the metrics were averaged over seven noise types (white noise, pink noise, speech noise, voice babble noise, F-16 noise, car factory noise and car Volvo-340 noise).

Fig. 17. The mean subjective preference score (%) comparison for each speech enhancement method. The female utterance (FF36_08) corrupted with 5 dB voice babble noise was used for the subjective tests. The error bars indicated the standard deviation of the scores.

Using the expression for the parabolic cylinder function from (26) and the relations [Th.9.212.2,9.212.3][38] yields Appendix D. MMSE-based Noise Power Estimation

Fig. 2. Mean preference scores (with standard error bars) for four stimuli types at: (a) —5 dB, (b) 0 dB, and (c) 5 dB Segmental SNR (SegSNR)

Fig. 3. Gain curves (with é = y — 1) describing: DCT MMSE spectral amplitude estimator with Gaussian prior Gy (é,y) defined by (15), indicated with solid line; the Ephraim and Malah solution [6] Gey (complex Gaussian speech prior), indicated with dashed line; Wiener filter solution Gy defined by (19) (Gaussian speech prior, linear filter), indicated with dash-dotted line. Nevertheless, the computation of (25) for a wide dynamic range is not trivial, and numerical problems may result when the arguments are large. To improve numerical stability, we rewrite (25) in terms of the modified Bessel functions (see Appendix C) spectral coefficient estimators always suppress more noise than the corresponding spectral amplitude estimators.

where ¢, =sgn(A_), I.,(Z) and K,(z) denote the modified Bessel functions of the first and second kind, respectively [eq.9.6.1] [42]. Similar to the Laplacian case, there is no closed-form solution for the equivalent DFT-based MMSE spectral amplitude estimator when the Gamma PDF is used [9,15]. Fig. 5 illustrates the gain

Fig. 4. Gain curves (with € = y — 1) describing: DCT MMSE spectral amplitude estimator with Laplacian prior G, defined by (22), indicated with solid line; DFT MMSE spectral amplitude estimator [9] G,_gs4 with v = 1 and K = 20 (i.e., the Taylor series of the modified Bessel function was truncated after 20 terms), indicated with dashed line; DCT MMSE spectral coefficient estimator with Laplacian speech prior [18] G,_csc, indicated with dashed-dotted line; DFT MMSE spectral coefficient estimator [14] G,_rsc with v = 1 and K = 20, indicated with dotted line. MATLAB implementations of the algorithms presented in [9,14] are available at [41].

Fig. 5. Gain curves (with € = y — 1) describing: DCT MMSE spectral amplitude estimator with Gamma prior G, defined by (27), indicated with solid line; DFT MMSE spectral amplitude estimator [9] Gc_gs, with v = 0.6 and K = 20 (i.e., the Taylor series of the modified Bessel function was truncated after 20 terms), indicated with dashed line; DCT MMSE spectral coefficient estimator with Gamma speech prior [18] Gc_csc, indicated with dashed-dotted line; DFT MMSE spectral coefficient estimator [14] Gc_psc with y = 0.6 and K = 20, indicated with dotted line. MATLAB implementations of the algorithms presented in [9,14] are available at [41].

Fig. 6. Gain curves plotted against the a priori SNR é and the instantaneous SNR y — 1 for the MMSE STSA estimators: (a) Gy (Gaussian speech prior) defined by (15), (b) G, (Laplacian speech prior) defined by (22), (c) Gc (Gamma speech prior) defined by (27), and (d) the Ephraim and Malah solution Ggy (complex Gaussian speech prior), as seen in [(14)] [6].

Fig. 7. Gain curves comparison for the proposed MMSE STSA estimators for (a) € = —15 dB, (b) € = —5 GB, (c) € = 5 dB and (d) € = 15 GB. The solid, dashed and dotted lines correspond to Gy (Gaussian speech prior), G, (Laplacian speech prior) and Gg (Gamma speech prior), respectively. The corresponding Wiener filter solution given by (19) (Gaussian speech prior, linear filter), and the Ephraim and Malah solution [(14)] [6] Gew (complex Gaussian speech prior), are respectively plotted with dash-dotted and loosely dotted lines for comparison.

Fig. 8. Gain curves for the MMSE STSA estimators incorporating speech presence uncertainty (SPU) with q = 0.2. (a) GY (Gaussian speech prior) defined by (36), (b) G’Y (Laplacian speech prior) defined by (37), (c) Gey (Gamma speech prior) defined by ()39, and (d) the respective Ephraim and Malah solution Gat (complex Gaussian speech prior), as seen in [(30)] [6].

Fig. 9. Gain curves for the proposed MMSE STSA estimators under speech presence uncertainty (SPU) for (a) € = —15 dB, (b) € = —5 dB, (c) € = 5 dB and (d) é = 15 GB, with q = 0.2. The solid, dashed and dotted lines correspond to Gj" (Gaussian speech prior) defined by (36), G}”" (Laplacian speech prior) defined by (37), and G2” (Gamma speech prior) defined by (39), respectively. The corresponding curves for the modified Wiener filter [11] (Gaussian speech prior, linear filter), and the Ephraim and Malah solution [(30)] [6] Gay ( (complex Gaussian speech prior), are respectively indicated with dash-dotted and loosely dotted line for reference.

Fig. 10. Performance comparison among various estimators tested using the oracle noise estimator (left column) and blind noise estimation given the noisy speech (righ column). Results are shown in terms of PESQ, STOI and Segmental SNR improvements and averaged over seven noise types (white noise, pink noise, speech noise, voice babbl noise, F-16 noise, car factory noise and car Volvo-340 noise).

Fig. 11. Performance comparison among various estimators with speech presence uncertainty (SPU) tested using the oracle noise estimator (left column) and blind nois¢ estimation given the noisy speech (right column). Results are shown in terms of PESQ, STO] and Segmental SNR improvements and averaged over seven noise types (white noise, pink noise, speech noise, voice babble noise, F-16 noise, car factory noise and car Volvo-340 noise).

Fig. 12. Spectrograms of (a) the clean sentence, (b) the sentence corrupted by non-stationary F-16 noise at O dB, and (c)-(j) enhanced speech produced by corresponding speech enhancement algorithm (see Table 1). The sentence 'Pitch the straw through the door of the stable’ (utterance MK62_09), was taken from the TSP speech database [47]. The nominated MMSE noise estimator introduced in Appendix D and in [35] were used for the DCT-based methods and DFT-based methods, respectively. The decision- directed approach [6] was used for the a priori SNR estimation.

Fig. 13. Spectrograms of (a) the clean sentence, (b) the sentence corrupted by voice babble noise at 0 dB, and (c)-(j) enhanced speech produced by corresponding speech enhancement algorithm (see Table 1). The sentence ‘The dune rose from the edge of the water’ (utterance FBO7_09), was taken from the TSP speech database [47]. The nominated MMSE noise estimator introduced in Appendix D and in [35] were used for the DCT-based methods and DFT-based methods, respectively. The decision-directed approach [6] was used for the a priori SNR estimation.

Fig. 14. The mean subjective preference score (%) comparison for each speech enhancement method. The male utterance (MK62_09) corrupted with 5 dB non-stationary F-" noise was used for the subjective tests. The error bars indicated the standard deviation of the scores.

Classification of the MMSE estimators with respect to transform domain. The noise is assumed to be additive and Gaussian. * indicate estimators for which no exact closed-form solutions exist. Table 1

Performance comparison, in terms of PESQ score gains, between various estimators tested using the nominated MMSE noise estimator.

Performance comparison, in terms of Segmental SNR gains, between various estimators tested using the nominated MMSE noise estimator.

Performance comparison, in terms of STOI score gains, between various estimators tested using the nominated MMSE noise estimator.

Noise estimator comparisons. p(Hi|Y) ~ 1, the noise update will cease and the noise estimate will remain the same as the previous frame’s estimate. Following the preceding computation of E{|DP\y}, the long-term noise PSD was then obtained via recursive smoothing with 8 = 0.95 p(Hi|Y) ~ 1, the noise update will cease and the noise estimate will remain the same as the previous frame’s estimate. Following the Cc a Oe tor, E{|DPIY, Ho}, with the periodogram of the noisy speech |Y|?. Furthermore, P(|Y||Ho) has the folded-normal PDF [37] Table 5 preceding computation of Ey IDI"IY +, the long-term noise PSD was

descriptionView Paper arrow_downwardDownload

Noise Variance Estimation for Spectrum Sensing in Cognitive Radio Networks

by adeel ahmed

2022, AASRI Procedia

descriptionView Paper arrow_downwardDownload

MobilDat-SK - a Mobile Telephone Extension to the SpeechDat-E SK Telephone Speech Database in Slovak

by Milan Rusko

2022

The paper describes design and process of collection, annotation and evaluation of a new Slovak mobile-telephone speech database MobilDat-SK, which is a mobile-telephone extension to the SpeechDat-E SK. The MobilDat-SK database contains... more

descriptionView Paper arrow_downwardDownload

From Semantic to Emotional Space

by Chew Lim Tan

2022

This paper proposes an effective approach to model the emotional space of words to infer their Sense Sentiment Similarity (SSS). SSS reflects the distance between the words regarding their senses and underlying sentiments. We propose a... more

descriptionView Paper arrow_downwardDownload

Cowles commission structural equation approach in light of nonstationary time series analysis

by Cheng Hsiao

2022

We review the advancement of nonstationary time series analysis from the perspective of Cowles Commission structural equation approach. We argue that despite the rich repertoire nonstationary time series analysis provides to analyze how... more

descriptionView Paper arrow_downwardDownload

Cowles commission structural equation approach in light of nonstationary time series analysis

by Cheng Hsiao

2022

descriptionView Paper arrow_downwardDownload

Statistical Methods for the Enhancement of Noisy Speech

by Jacob Benesty

2022, Signals and Communication Technology

With the advent and wide dissemination of mobile communications, speech processing systems must be made robust with respect to environmental noise. In fact, the performance of speech coders or speech recognition systems is degraded when... more

descriptionView Paper arrow_downwardDownload

Generalization of pre-image iterations for speech enhancement

by Franz Pernkopf

2021, 2013 IEEE International Conference on Acoustics, Speech and Signal Processing

In this paper, we extend the pre-image iteration method for speech de-noising by automatic determination of the kernel variance. The kernel variance needs to be adapted in different noise conditions. In previous work, the signal-to-noise... more

descriptionView Paper arrow_downwardDownload

Phone-Conditioned Suboptimal Wiener Filtering

by Doroteo Toledano

2021, 2010 20th International Conference on Pattern Recognition

A novel way of managing the compromise between noise reduction and speech distortion in Wiener filters is presented. It is based on adjusting the amount of noise reduced, and therefore the speech distortion introduced, on a phone-by-phone... more

descriptionView Paper arrow_downwardDownload

Phone-Conditioned Suboptimal Wiener Filtering

by Doroteo Toledano

2021, 2010 20th International Conference on Pattern Recognition

descriptionView Paper arrow_downwardDownload

Impact by Intention: An Argument for Forensics as a High Impact Practice

by Vincent L . Stephens

2020, National Forensic Journal

This essay locates forensics within national discourse about high-impact practices (HIPs) in higher education, as outlined by scholar George D. Kuh. Forensics shares all the characteristics associated with the ten promising practices Kuh... more

descriptionView Paper arrow_downwardDownload

Example of Graduation Speech

by Suhada F. Mahdi

2020

descriptionView Paper arrow_downwardDownload

Implementation of Bayesian Recursive State-Space Kalman Filter for Noise Reduction of Speech Signal

by Ali Sarafnia

2017

—Noise reduction of speech signals plays an important role in telecommunication systems. Various types of speech additive noise can be introduced such as babble, crowd, large city, and highway which are the main factor of degradation in... more

descriptionView Paper arrow_downwardDownload

Noise Reduction of Speech Signal Using Bayesian State-Space Kalman Filter

by Ali Sarafnia

2017

— The noise exists in almost all environments such as cellular mobile telephone systems. Various types of noise can be introduced such as speech additive noise which is the main factor of degradation in perceived speech quality. At some... more

descriptionView Paper arrow_downwardDownload

The Employment of Bayesian Method in Noise Reduction and Packet Loss Replacement

by Ali Sarafnia and

2017

Speech enhancement in real-time applications improves the quality and intelligibility of the speech and reduces communication fatigue. Nowadays, due to reactivity of the systems and spread of online real-time applications, including VoIP,... more

descriptionView Paper arrow_downwardDownload

Why Didn't You Talk to Your Mommy, Honey?": Parents' and Children's Talk About Talk

by Jean Berko Gleason and

2016, Research on Language & Social Interaction

descriptionView Paper arrow_downwardDownload

Acknowledgments

by Delinda Mercer

2015, Speech Communication

descriptionView Paper arrow_downwardDownload

Learning speech features in the presence of noise: Sparse convolutive robust non-negative matrix factorization

by Scott Rickard

2015, 2009 16th International Conference on Digital Signal Processing

We introduce a non-negative matrix factorization technique which learns speech features with temporal extent in the presence of non-stationary noise. Our proposed technique, namely Sparse convolutive robust non-negative matrix... more

The SCRNMF estimate, Er, of the data matrix, V, has the form, We consider an extended K LD objective function in (15) and de- rive multiplicative diagonally rescaled gradient descent update rules using the model (13). We constrain the basis, W, to have unit Lo norm so that the sparsity constraint on speech portion of H is en-

We propose the new extended convolutive basis [W,, I]. This basis is partitioned such that only the speaker portion, W,, is up- dated. W denotes temporal slice ¢ and each atom of the basis has been normalized atom-wise as in [9]. We model the noise as a recti- fied Gaussian random variable in each time-frequency bin using the basis I. For example, for a factorization of rank R and an observa- tion matrix V of dimension M x N the SCRNMF activation matrix has the structure,

Fig. 1. Illustrating how the a-divergence family of objective func- tions and Squared Euclidean Distance (SED) penalize error. The objective value is plotted as a function of the estimate. The solu- tion is 3. SED penalizes under and over estimation equally. The a- divergence family penalizes under estimation more harshly than over estimation which motivates its use for decomposing magnitude spec- trograms of speech as the formants of speech manifest themselves as peaks in the magnitude spectrogram. The formants of speech are of particular interest in speech feature learning.

Fig. 2. Trade-off between sparsity of the activations and accuracy of the reconstruction. We plot the sparsity of the activation matrix H as function of reconstruction error for a range of values of \ using the SCNMF with the a-divergence family of objective functions. For each set of parameters, e.g. a = 0.5, KLD and a = 1.5, we leam a decomposition of 80 and 100 atoms which we indicate in paren- thesis in the legend. Increasing the number of atoms from 80 to 100 atoms in each case reduces the reconstruction error for a given level of sparsity. Setting a = 0.5 yields the decomposition with the worst reconstruction error fora given level of sparsity. The KLD objective gives a middle ground performance between a = 0.5 anda = 1.5 with respect to sparsity of H and reducing the reconstruction er- ror. We use this plot as a guide to choosing the parameter in the SCRNMF objective (15) in subsequent experiments.

Fig. 3. Trade- off between de-noising parameter 3 and the accuracy of the extracted speech using the SCRNMF algorithm with the KLD objective (15). For synthetically generated mixtures with SNRs of 5,0, —3dB and a value for \ chosen for the SCRNMF KLD objec- positions of ran’ composition and tive in Fig 2, we tune the de-noising parameter 3. We learn decom- k 80 and 100 atoms. Each generated curve using SCRNMF is labeled using the algorithm name, the rank of the de- the mixture SNR, e.g. SCRNMF 100(5dB). The benchmark is the SCNMF algorithm which is a straight line (for each 3) and is indicated by the label KLD (at each SNR level). The SC- MF objective best result from over a range of 9) does not depend on the 3 parameter. We plot the the 80 and 100 atom decomposition for SCNMF as the technique performs similarly in each case and the two resulting ines would be indistinguishable. We achieve gains in performance values of G over SCNMF using SCRNMF. As the evel of noise added to the mixture decreases the performance gain decreases. For example, for a mixture of SNR —3dB we achieve a gain of ~ 3dB over SCNMF and for a mixture of SNR 5dB we im- prove the solution by ~ 1.5dB. This is expected as with oodB we expect both SCNMF and SCRNMF to perform similarly. We mea- sure performance by comparing the estimated de-noised speech with he clean speech. Increasing the number of atoms at high mixture SNRs improves the performance.

Table 2. Comparision of enhancement techniques in various noisy conditions.

descriptionView Paper arrow_downwardDownload

Understanding the significance of radiometric calibration for synthetic aperture radar imagery

by Eric Gill

2015, 2014 IEEE 27th Canadian Conference on Electrical and Computer Engineering (CCECE)

In applications such as target recognition, quantitative use of the information present in synthetic aperture radar (SAR) imagery is pivotal for detecting and classifying the scattering centers of the target(s). This paper presents an... more

descriptionView Paper arrow_downwardDownload

Effect of detection on spatial resolution in synthetic aperture radar imagery and mitigation through upsampling

by Eric Gill

2015, Journal of Applied Remote Sensing

The complex-valued image output from a synthetic aperture radar (SAR) processor possesses full spatial resolution defined by the sensor. Typically, this image is either power detected or magnitude detected before it is subjected to... more

descriptionView Paper arrow_downwardDownload

Securing Speech in GSM Networks using DES with Random Permutation and Inversion Algorithm

by Call for paper-International Journal of Distributed and Parallel Systems (IJDPS) and

2015

descriptionView Paper arrow_downwardDownload

Secure Data and Voice Transmission over GSM Voice Channel: Applications for Secure Communications

by Aldo F Dragoni

2014

The GSM voice channel is the world's most widely used mobile communication network. Unfortunately these networks are affected by serious vulnerability from hardware-based attacks and communications can be easy to intercept. This paper... more

descriptionView Paper arrow_downwardDownload

Effect of detection on spatial resolution in synthetic aperture radar imagery and mitigation through upsampling

by Khalid El-Darymli

2014

descriptionView Paper arrow_downwardDownload

Outline Gaymarriages

by Tyrel Brown

2013

Specific Purpose: The specific purpose of this speech is to educate people about marriages and persuade people to help in the fight towards legalizing people of the same sex to be married and not put on contract.

descriptionView Paper arrow_downwardDownload

Understanding the Significance of Radiometric Calibration for Synthetic Aperture Radar Imagery

by Khalid El-Darymli and

2012, IEEE Canadian Conference on Computer and Electrical Engineering'14