DSpace Community:http://localhost:8081/xmlui/handle/123456789/82024-01-07T08:32:59Z2024-01-07T08:32:59ZDESIGN AND EVALUATION OF ENHANCEMENT TECHNIQUES FOR SINGLE-CHANNEL SPEECHSingh, Sachinhttp://localhost:8081/xmlui/handle/123456789/154772023-06-16T12:00:53Z2015-04-01T00:00:00ZTitle: DESIGN AND EVALUATION OF ENHANCEMENT TECHNIQUES FOR SINGLE-CHANNEL SPEECH
Authors: Singh, Sachin
Abstract: In real-world applications, a speech signal from the uncontrolled environment is often
accompanied by various degradation components along with the actual speech components.
- The degradation components include background noise, reverberation and multi-talker
speech. These unwanted interferences not only degrade perceptual speech quality and
intelligibility which creates listening problem for human, but also give poor performance in
automatic speech processing tasks like speech recognition, speaker recognition and hearing
aid systems. Therefore, de-noising of corrupted single-channel speech has become a very
necessary and important aspect for research in academia and industry.
The presently available single channel noise reduction methods include spectral
subtraction, Wiener filter, minimum mean square error estimation (MMSE) and p-MMSE,
log-MMSE, KLT, PKLT etc. These methods are applicable for specific environment of
speech signal. Some of these perform better for one particular types of noise whereas others
are suitable for other types of noise. Considering the limitations of these methods, different
categories of speech signals have been treated separately. Based on this, the objectives of the
- present research work have been formulated as: (1) design of a suitable method for
enhancement of mixed noisy speech of very low (Negative) input SNR conditions; (2) design
and development of a suitable method for suppression of non-stationary noise in singlechannel
speech signal; (3) analysis and development of a suitable method for suppression of
combined effect of background noise and reverberation; and (4) design and implementation of
phase based single-channel speech enhancement technique. The mentioned objectives have
been accomplished as follows:
In the first objective, single-channel speech enhancement based on modified Wiener gain
function using Wavelet Packet Transform (WPT) is proposed for suppression of noise from
multiple sources in both the low (negative) and high SNR speech signal ranging from -15 dB
to +15 dB. The method includes steps as (1) decomposition of speech signal upto 3 d level to
get speech signal in eight different bands; (2) the FFT of these bands is computed to get the
wavelet packet soft threshold which is applied on the above FFT output; (3) the WP soft
threshold is also used to determine the modified gain function; (4) finally to get the processed
output speech, the IFFT of the product of the modified Wiener gain function and WP
thresholded FFT output is computed. The overlap-add method is used to get the end
reconstructed speech signal.
1
The performance of this proposed method is compared with other existing speech
enhancement methods evaluating their performance parameters such as MSE, SNR, MOS,
PESQ and SII. The dataset of low SNR ranging from -15 dB to +15 dB having mixed noise is
used for performance evaluation of the implemented methods. The results show the
improvement in terms of speech quality and intelligibility parameters. Proposed method gives
highest improvement in comparison to other single-channel speech enhancement methods for
all input SNR levels with various noise types.
To overcome the problem of using true speech or true noise in binary mask based methods
of speech enhancement, a fuzzy mask is proposed here under second objective. It is based on
soft and hard wavelet packet threshold. The method includes steps as (1) decomposition of
speech signal upto 3'' level to get speech signal in eight different bands; (2) the FFT of these
bands is computed to get the wavelet packet soft and hard threshold which is applied on the
FFT output; (3) in this procedure, the modified Wiener gain function determined similarly as
above is applied to get the denoised speech signal in frequency domain at first stage; (4) in
second stage, fuzzy mask is applied on the output of first stage for further enhancement; (5)
finally to get the processed output speech, the IFFT of the product of the fuzzy mask and WP
soft and hard thresholded FFT output is computed. Again, the overlap-add method is used to
get the end reconstructed speech signal.
Here again, the performance of this proposed method is compared with other existing
speech enhancement methods comparing their performance parameters such as SNR, MOS,
PESQ and STOI. The dataset of low SNR ranging from -15 dB to +15 dB having nonstationary
noise is used for performance evaluation of the implemented methods. The results
obtained from proposed method are much better than other existing single-channel speech
enhancement methods.
Most of the above implemented algorithms are used for speech enhancement of noise and
reverberation separately and they do not work effectively in case of their combination (i.e.
reverberation with noise). To suppress the combined effect of early and late reverberations
with various types of noise, a binary reverberation mask is implemented here for the
fulfillment of the third objective.
In this proposed method signal-to-reverberant ratio (SRR) is calculated as a limit for ideal
reverberant mask (IRM). The amplitudes with SRR greater than a preset threshold (i.e. -5dB)
are used for reconstruction of dereverberated speech, while amplitudes with SRR values
smaller than the threshold are eliminated. The construction of the SRR criterion assumes a
priori knowledge of the input reverberant and target signal. Threshold values varying from
0dB to -90dB are analyzed for selection of IRM limit T. Finally, the dereverberated speech
signal is constructed by multiplying noisy speech with reverberant mask.
The proposed reverberant mask based speech enhancement method is compared with
other existing speech enhancement methods in terms of speech quality and intelligibility
measure parameters such as PESQ, CD, SNR and MSE. The maximum improvement in
reverberated noisy speech is obtained by proposed method in terms of speech quality and
intelligibility at all input SNR levels ranging from -25 dB to -5 dB.
Most of the noise reduction algorithms perform the modification in amplitude only, while
phase remains unchanged or discarded in the process of speech enhancement. Recently, it has
been found that quality and intelligibility both can be improved upto a significant level by
using either phase of speech signal only or phase with amplitude. Hence, signal phase ratio
based single-channel speech enhancement method is implemented in fourth objective for
further improvement in noisy speech signal which considers the phase of the noisy speech
signal in processing. The phase ratio of noisy speech to noise signal is used in the phase based
method. In this method two gain functions Gi and G2 are developed for correction in noisy
phase by suppressing the noise coming from angles between 0 to ±7t/2 and ±2r/2 to ±ir,
respectively. For the reconstruction of speech spectrum, both gains are multiplied together
and lower values of the phases are neglected for getting desired speech spectrum. Results are
compared with other phase based methods (such as phase spectrum compensation (PSC),
exploiting conjugate symmetry of the short-time Fourier spectrum and STFT-phase for the
MMSE-optimal) and are analyzed in terms of speech quality, intelligibility measures (like
SNR, SSNR, SIG, SII, BAK, OVL, and PESQ, etc.), informal subjective listening tests and
spectrogram analysis. The performance measure parameters show that the proposed phase
ratio based implemented method provides more effective improvement in noisy speech in
comparison to other phase based speech enhancement methods.
Implemented algorithms are evaluated for various languages i.e. Hindi, Kannada, Bengali,
Malayalam, Tamil, Telgu, and English. Indian language database used for evaluation are
taken from lilT-H Indic Speech Databases which was developed at Speech and Vision Lab,
IiIT-Hyderabad for the purpose of building speech synthesis system among Indian languages.
The speech data were recorded by native speakers of each language. The recording was done
in a studio environment using a standard headset microphone connected to a Zoom handy
recorder. A set of 1000 sentences were selected for each language. These sentences were
selected to cover 5000 most frequent words in text corpus of the corresponding language. The
NOIZEUS database of clean and noisy speech was used for English language sentences. This
database basically contains 30 IEEE sentences which were produced by three male and three
female speakers in groups. The real-world sources of background noise at different SNRs
were taken from AURORA and NOISEX-92 databases, respectively which include suburban
train noise, babble, car, exhibition hall, restaurant, street, airport and train-station as noise
sources.
In the nut shell, it can be said that the present work is an effort to determine suitability of
various single-channel speech enhancement techniques to get the maximum speech quality
and intelligibility.2015-04-01T00:00:00ZDEVELOPMENT OF PROTECTION ALGORITHMS FOR COMPENSATED TRANSMISSION LINEManori, Ashokhttp://localhost:8081/xmlui/handle/123456789/154762023-06-16T12:00:12Z2016-05-01T00:00:00ZTitle: DEVELOPMENT OF PROTECTION ALGORITHMS FOR COMPENSATED TRANSMISSION LINE
Authors: Manori, Ashok
Abstract: Flexible Alternating Current Transmission System (FACTS) controllers can be
connected throughout the complete length of transmission line. Moreover, the preferred
location is at the middle of the line because voltage fluctuation is maximum at that point.
Particularly, shunt FACTS controllers are preferably connected in the middle as they provide
voltage support to the line. When FACTS controllers are connected at the sending or receiving
end substations the protection of transmission line becomes somewhat easier because
measurements can be taken locally without any considerable time delay. If the FACTS
controller is placed at the middle of transmission line, the protection of transmission line
becomes a challenging task for protection engineers. In such condition transmission line has
two sections: one before the FACTS device which is uncompensated portion whose impedance
is linear and other after the device which is being compensated and it's having variable
impedance (non-linear). Therefore, ultimate challenge for protection engineers is to protect
such a line whose impedance is not linear throughout the whole length. To protect transmission
line having midpoint compensation, first requirement is to know the section of line in which
fault occurred. After identifying fault section, an adaptive protection algorithm is needed to
protect the complete line. In this work, fault section is identified by a combined wavelet-SVM
based technique which takes very less time to analyze the pattern and select the faulty section.
Wavelet transform decomposes all the frequency bands present in fault signal and SVM
analyze the frequency patterns for different fault sections. Any transmission line fault signal
has various frequency signals up-to 80 kHz. Relay, which is placed at the bus, receive all the
frequency signals including fundamental frequency signal through current transformer. Any
fault signal occurred before or after the FACTS device has different frequency spectrum
because of the filtering by FACTS device itself. FACTS device filters the particular frequency
band from the fault signal according to its inductance or capacitance value at that time (or
according to its compensation level). In case of series compensation, inductance or capacitance
value inserted by FACTS device has been used as filtering element, therefore, it was claimed
that particular frequency signal will be missing if fault occurred in section-Il. Although, in case
of shunt compensation like SVC and STATCOM, impedance inserted by the device is not
constant, so it cannot be said surely that which frequency band will be missing in frequency
spectrum. Therefore, to make transmission line protection system reliable, it is needed to find
fault section clearly. For this purpose, a high pass filter is installed at the FACTS device
location which surely filters the high frequency signals generated in section two in case of fault.
In case of fault in section one, wavelet transform provides detail about high frequency
components, while, fault in section two has no or less detail about high frequency signals. One
drawback of the wavelet transform is its sensitivity for noise signals. Therefore, to make relay
performance less prone to noise signal, a SVM classifier is added to make protection algorithm
more adaptive. SVM can draw an optimal hyper plane to categorize two classes, therefore, it
can classify the wavelet signals optimally for section one or two faults. Further, SVM
parameters are optimized by Genetic Algorithm to make classifier an optimal classifier. Results
are compared with the RBFNN based classifier.
Presence of TCSC in the middle of transmission line creates major protection issues
particularly in distance protection. First section of line has linear impedance trajectory and can
be easily protected by the convention mho relay algorithm. After TCSC, point impedance of
line is a shifted trajectory. In capacitive mode of operation, it shifts into downward direction
and in inductive mode of operation it shifts toward upward direction and in both the cases
amount of shift depends upon the level of compensation. Therefore, requirement of the
transmission line protection system is such that it should be adaptive with the shifted
impedance trajectory of the second portion of transmission line. Another challenge in
protection of transmission line having TCSC is that, TCSC itself protected by MOV varistor
which is a non-linear resistor. When fault current exceeds a certain limit, overvoltage across
TCSC capacitor become too high so, to protect TCSC from overvoltage, MOV discharge
excessive energy through itself. In this case the equivalent impedance of TCSC with MOV is
not linear and measured fault impedance has uncertain error. It is the requirement of the power
engineer to place TCSC safely in the transmission line even in fault condition such that as soon
as fault cleared it should function on the line accordingly to reduce economic losses. Therefore,
in any short duration fault or transient condition, TCSC remain in the line and inserts its own
impedance. To make transmission line protection system adaptive, it is required to calculate
TCSC impedance and with respect to this impedance it is required to make transmission line
protection system adjustable. TCSC equivalent impedance depends on firing angle of thyristor,
therefore, it can be calculated by measuring a single variable. MOV setting is kept such that it
operates in highly severe condition and bypasses high amount of current. In this case, TCSC
and MOV equivalent impedance comes into picture.
ii
Similarly, the presence of shunt FACTS device such as SVC and STATCOM also
creates some major protection issues when placed at the middle of transmission line. Certainly,
the preferred location for shunt FACTS device is at the middle because at this point voltage sag
is maximum. Shunt devices provide voltage support at the point of connection by injecting a
quadrature current into the line. Shunt devices does not affect the performance of relay for any
fault before the point of connection. For any fault after the point of connection it introduces
certain error in fault impedance calculation according to current injected by the device. Shunt
devices inject both the leading or lagging current at the point of common coupling depending
on the system requirement hence creates under and over reaching problems. Therefore, to make
transmission line protection system adaptable with the action of shunt FACTS device it is
required to calculate the error introduced by the device. A compensation unit has been inserted
in the conventional Mho relay which calculates error in terms of current injected by the shunt
device into the transmission line.
Fault zone identification in compensated transmission line is constrained by several
factors like fault resistance, loading conditions and mainly by the level of compensation at the
time of fault. For a secure transmission line protection system, it is needed to deal with all
above mentioned factors when deciding the faulty zone. In uncompensated transmission line,
some fault zone identification techniques are proposed which make transmission line protection
system adaptable with loading conditions and different values of fault resistance. In these zone
identification techniques, fault impedance coordinates have been calculated for different types
of faults and power system operating conditions. Decision is taken by training any intelligence
tool like SVM to decide the particular fault zone. These techniques are adaptive to changing
loading conditions and various values of fault resistance. In case of compensated transmission
line, inductive portion of transmission line is also variable and can emerge in any direction
depending on the mode of operation of the compensator. Fault impedance coordinates of
different zones for these lines are overlapping over the imaginary axis at the intersection of
zones. Even the learning of intelligence tool by these coordinates cannot solve the problem
completely because for two different fault locations same coordinates may exist. So, to solve
zone identification problem in compensated transmission line, work is needed to be done on
both the sides; first on fault impedance coordinates and second on intelligence tool. In this
work, impedance inserted by compensator into the transmission line is estimated and modified
impedance coordinates of fault impedance is calculated for the learning of intelligence tool.
111
Due to several advantageous features, SVM is used for fault zone identification in compensated
transmission line. Results are compared with the RBFNN based classifier.2016-05-01T00:00:00ZFEATURE EXTRACTION AND CLASSIFICATION OF EEG SIGNALSM., Nikhilhttp://localhost:8081/xmlui/handle/123456789/154222022-06-03T06:14:02Z2013-06-01T00:00:00ZTitle: FEATURE EXTRACTION AND CLASSIFICATION OF EEG SIGNALS
Authors: M., Nikhil
Abstract: The EEG biometric system will be superior in performance and much reliable. In the case of EEG based security system, the mere presence of a person is not enough to bypass the system. EEG of same person will differ with the state of the person and conditions under which it is recorded. The biometric security system will have a sample of EEG recorded under some standard condition. So if a person is forcefully asked to bypass the system, he will not be able to bypass it. This is not the case with most conventional security systems like fingerprinting. Here we may be able to bypass the system, by forceful placement of finger printing or using some fake finger print. The challenging part in this will be to identify and calculate the features of EEG which will give best results. This is a case dependent problem, and hence we cannot say for certain that a particular set of features will give best result. Further the classification efficiency will also depend up on the type of classifier used. in my thesis work I tried to classify EEG belonging to different persons and thereby there by trying to classify different person. If this classification works well for a large number of persons of whom some may be intruders i.e., their EEG is not in the list of EEG patterns of persons whom we have to identify, then we will be able to make a biometric security system.2013-06-01T00:00:00ZAN ADAPTIVE POWER LINE INTERFERENCE CANCELLER FOR ELECTROCARDIOGRAPHYMishra, Saritahttp://localhost:8081/xmlui/handle/123456789/154212022-06-03T06:12:49Z2013-06-01T00:00:00ZTitle: AN ADAPTIVE POWER LINE INTERFERENCE CANCELLER FOR ELECTROCARDIOGRAPHY
Authors: Mishra, Sarita
Abstract: A Gocrtzcl based all digital phase locked loop for removing power line interference
from the ECG signal is proposed. A power line interference is artificially
iiitroduccd in ECG signal and this proposed scheme tracks the amplitude and
phase of all the interference components for power line frequency deviations from
input signal frequency (i.e. 50 Hz). The output sampling frequency is adjusted
with the help of all digital phase locked loop (ADPLL) to correct the phase errors
introduced by the SG filter whenever variations in frequency occur. Main
building blocks of proposed method are discussed. MATLAB simulation results
are discussed and future work is to make this scheme adaptive for large variation
in input frequency.2013-06-01T00:00:00Z