Electronics and Communication Engineering

This paper presents an adaptive speech enhancement approach to suppress non-stationary noises form a
noisy speech signals. This approach is based on the Empirical Mode Decomposition and Signal Uncertainty. The
EMD is just used as pre-processor for signal decomposition into Intrinsic Mode Functions. Further The IMFs are
processed for noise suppression through a recursive smoothing based on a smoothing factor which was decided
based on the probability of speech presence. A new signal detector is proposed here to measure the probability of
speech presence. This approach mainly focused on the optimization of probability of detection through the newly
proposed signal detector. The simulation is carried out through various speech signals contaminated with different
noise types like White noise, Babble noise and Airport noise at various SNR levels reveals the outstanding
performance of proposed approach. The performance evaluation is carried out by measuring the performance metrics,
Overall output SNR, Output AvgSegSNR and Perceptual Evaluation of Speech Quality (PESQ). The evaluation is
carried out for varying noise strengths and for every test case all these metrics are evaluated and compared with
conventional approaches.
Keywords: Speech enhancement, Non-stationary noise, Speech uncertainty, EMD, PESQ.

This paper proposes a new speech enhancement framework to improve the quality of speeches recorded under adverse acoustic
environments based on the speech presence uncertainty. Since the uncertainty evaluation gives a more and clear discrimination about the
speech and noise, this paper proposes a new uncertainty evaluation mechanism as a preprocessing mechanism to the noise suppression
methods. This mechanism relates with energies of a noisy speech signal and classifies the speech segments and noise segments more
perfectly. In addition to the quality enhancement, this approach also reduces the unnecessary computational burden over the speech
processing system. Extensive simulations are carried out over the speech signals with different types of non-stationary noises like babble
noise, exhibition noise, restaurant noise and train station noises and the performance is measured with the performance metrics namely
the Output SNR, AvgSegSNR, PESQ and COMP. The comparative analysis of proposed approach over the conventional approaches
shows an outstanding performance in all environments.
Keywords: Speech enhancement, Noise estimation, EMD, Thresholding, Babble noise, Output SNR, and PESQ

Abstract: Suppression of non-stationary noise from a noisy speech is very important for speech oriented applications like speech

recognition, speech communication etc. This paper proposes a new speech enhancement approach to suppress the non-
stationary noises like Babble noise, car noise etc., based on Discrete Cosine Transform and Empirical Mode Decomposition. The

complete accomplishment of proposed approach is carried out in two phases, DCT based soft thres holding and Adaptive
Empirical mode Decomposition based denoising. Initially the noise speech is filtered through DCT based soft thres holding and
the partially enhanced speech is processed through Adaptive Empirical mode Decomposition to reduce the residual noise further.
An adaptive noise cancellation procedure is accomplished which considers the correlation between the noise and the reference signal to
filter the noise. The extensive simulations are carried out over the proposed approach through different speech signals with
different noises at different signal strengths and the performance is evaluated through Output SNR, Output Avg Seg SNR and
PESQ test.
Keywords: Speech enhancement, Discrete Cosine Transform, Empirical Mode Decomposition, Babble Noise, Output SNR,
PESQ.

Generally, Speech enhancement aims to improve speech quality and intelligibility of a noise
contaminated speech signal by using various signal processing approaches. Removal of a noise from a noisy
speech is a common problem; already a vast research was carried out in earlier. However, due to the
characteristics of various types of noises, the approaches proposed in earlier are not applicable for all types of
noises. In addition, the earlier approaches didn’t focus on the non-linear and non-stationary characteristics on
noise environments. EMD is a filtering approach performs efficiently for non-stationary environments. This
paper proposes a novel EMDF approach with the inspiration of thresholding to remove the noise from noisy
speech sample. The proposed approach also developed a method to select the IMF index for separating the
residual low-frequency noise components from the speech estimate, based on the IMF statistics. An
experimental study was also done on various types of noise contaminated speech samples like babble noise,
restaurant noise and car interior noise at various strengths.
Keywords: Speech enhancement, EMDF, IMF, noise estimation, SegSNR.

A new method for the enhancement of speech signals
contaminated by speech-correlated noise, such as that
in the output of a speech coder, is presented. This
module is based on numerical speech processing
algorithms which modelise the infected ear and
generates the stimulus signals for the cilia cells
(brain). The method is also based on constrained
optimization of a criterion. This interface uses a
gamma chirp filter bank constituted of 16 band pass
filters based on IIR filters. The implemented method
is on a block by- block basis and uses two constraints.
A first constraint ensures that the signal power is
preserved. A modification constraint ensures that the
power of the difference of the enhanced and
unenhanced signal is less than a fraction of the
power of the unenhanced signal. The applied method
is to increase the periodicity of the speech signal.
Sounds that are not nearly periodic are perceptually
unaffected by the optimization because of the
modification constraint. The results demonstrated a
degree of discrimination and interferences between
different sounds especially in multi speaker
environment.
Ke

Abstract—This paper proposed an adaptive hybrid speech
enhancement algorithm to reduce the non-stationary noise of a
noise contaminated speech. This approach combines the Discrete
Cosine Transform (DCT) with an Adaptive Empirical Mode
Decomposition (AEMD). The proposed DCT is an extension to
the general DCT and the AEMD was developed by adding an
extra noise cancellation block for enhancement. The complete
approach was carried out in two stages. In the first stage, the
noise contaminated speech was enhanced through DCT and
adaptive soft thresholding. In second stage, to further reduce the
residual noise present in enhanced speech, AEMD was applied
through an adaptive noise cancellation block. The performance
evaluation of proposed approach was under white and babble
noises at various SNR levels.
Index Terms—Speech Enhancement, DCT, EMD,
thresholding, IMF, Non-stationary noise.

Abstract: In this paper, a novel speech enhancement approach
was proposed to improve the quality of the speech contaminated
with various types of non-stationary noises. An EMD based clear
recursive thresholding (EMD-CRT) approach was proposed in
this approach, inspired by wavelet thresholding. This approach
performs the thresholding operation on the noisy speech
recursively, such that the non-stationary noises are getting
filtered effectively. To evaluate the performance, the proposed
approach was tested under various levels of SNR and also for
various types of non-stationary noises like babble noise, car
noise, restaurant noise and airport noise. A comparative analysis
was also done with earlier approach to check the performance of
the proposed approach. A subjective speech quality evaluation
was also done by performing the mean opinion score (MOS).
Keywords: Speech enhancement, non-stationary noises, EMD,
thresholding, SegSNR, RER, MOS.