Abstract: This paper presents a formant-tracking linear prediction
(FTLP) model for speech processing in noise. The main focus of this
work is the detection of formant trajectory based on Hidden Markov
Models (HMM), for improved formant estimation in noise. The
approach proposed in this paper provides a systematic framework for
modelling and utilization of a time- sequence of peaks which satisfies
continuity constraints on parameter; the within peaks are modelled
by the LP parameters. The formant tracking LP model estimation
is composed of three stages: (1) a pre-cleaning multi-band spectral
subtraction stage to reduce the effect of residue noise on formants
(2) estimation stage where an initial estimate of the LP model of
speech for each frame is obtained (3) a formant classification using
probability models of formants and Viterbi-decoders. The evaluation
results for the estimation of the formant tracking LP model tested
in Gaussian white noise background, demonstrate that the proposed
combination of the initial noise reduction stage with formant tracking
and LPC variable order analysis, results in a significant reduction in
errors and distortions. The performance was evaluated with noisy
natual vowels extracted from international french and English vocabulary
speech signals at SNR value of 10dB. In each case, the
estimated formants are compared to reference formants.
Abstract: The speech signal conveys information about the
identity of the speaker. The area of speaker identification is
concerned with extracting the identity of the person speaking the
utterance. As speech interaction with computers becomes more
pervasive in activities such as the telephone, financial transactions
and information retrieval from speech databases, the utility of
automatically identifying a speaker is based solely on vocal
characteristic. This paper emphasizes on text dependent speaker
identification, which deals with detecting a particular speaker from a
known population. The system prompts the user to provide speech
utterance. System identifies the user by comparing the codebook of
speech utterance with those of the stored in the database and lists,
which contain the most likely speakers, could have given that speech
utterance. The speech signal is recorded for N speakers further the
features are extracted. Feature extraction is done by means of LPC
coefficients, calculating AMDF, and DFT. The neural network is
trained by applying these features as input parameters. The features
are stored in templates for further comparison. The features for the
speaker who has to be identified are extracted and compared with the
stored templates using Back Propogation Algorithm. Here, the
trained network corresponds to the output; the input is the extracted
features of the speaker to be identified. The network does the weight
adjustment and the best match is found to identify the speaker. The
number of epochs required to get the target decides the network
performance.
Abstract: Detection and tracking of the lip contour is an important
issue in speechreading. While there are solutions for lip tracking
once a good contour initialization in the first frame is available,
the problem of finding such a good initialization is not yet solved
automatically, but done manually. We have developed a new tracking
solution for lip contour detection using only few landmarks (15
to 25) and applying the well known Active Shape Models (ASM).
The proposed method is a new LMS-like adaptive scheme based on
an Auto regressive (AR) model that has been fit on the landmark
variations in successive video frames. Moreover, we propose an extra
motion compensation model to address more general cases in lip
tracking. Computer simulations demonstrate a fair match between
the true and the estimated spatial pixels. Significant improvements
related to the well known LMS approach has been obtained via a
defined Frobenius norm index.
Abstract: The transformation of vocal characteristics aims at
modifying voice such that the intelligibility of aphonic voice is
increased or the voice characteristics of a speaker (source speaker) to
be perceived as if another speaker (target speaker) had uttered it. In
this paper, the current state-of-the-art voice characteristics
transformation methodology is reviewed. Special emphasis is placed
on voice transformation methodology and issues for improving the
transformed speech quality in intelligibility and naturalness are
discussed. In particular, it is suggested to use the modulation theory
of speech as a base for research on high quality voice transformation.
This approach allows one to separate linguistic, expressive, organic
and perspective information of speech, based on an analysis of how
they are fused when speech is produced. Therefore, this theory
provides the fundamentals not only for manipulating non-linguistic,
extra-/paralinguistic and intra-linguistic variables for voice
transformation, but also for paving the way for easily transposing the
existing voice transformation methods to emotion-related voice
quality transformation and speaking style transformation. From the
perspectives of human speech production and perception, the popular
voice transformation techniques are described and classified them
based on the underlying principles either from the speech production
or perception mechanisms or from both. In addition, the advantages
and limitations of voice transformation techniques and the
experimental manipulation of vocal cues are discussed through
examples from past and present research. Finally, a conclusion and
road map are pointed out for more natural voice transformation
algorithms in the future.
Abstract: In this paper in consideration of each available
techniques deficiencies for speech recognition, an advanced method
is presented that-s able to classify speech signals with the high
accuracy (98%) at the minimum time. In the presented method, first,
the recorded signal is preprocessed that this section includes
denoising with Mels Frequency Cepstral Analysis and feature
extraction using discrete wavelet transform (DWT) coefficients; Then
these features are fed to Multilayer Perceptron (MLP) network for
classification. Finally, after training of neural network effective
features are selected with UTA algorithm.
Abstract: Preprocessing of speech signals is considered a crucial step in the development of a robust and efficient speech or speaker recognition system. In this paper, we present some popular statistical outlier-detection based strategies to segregate the silence/unvoiced part of the speech signal from the voiced portion. The proposed methods are based on the utilization of the 3 σ edit rule, and the Hampel Identifier which are compared with the conventional techniques: (i) short-time energy (STE) based methods, and (ii) distribution based methods. The results obtained after applying the proposed strategies on some test voice signals are encouraging.
Abstract: A talking head system (THS) is presented to animate
the face of a speaking 3D avatar in such a way that it realistically
pronounces the given Korean text. The proposed system consists of
SAPI compliant text-to-speech (TTS) engine and MPEG-4 compliant
face animation generator. The input to the THS is a unicode text that is
to be spoken with synchronized lip shape. The TTS engine generates a
phoneme sequence with their duration and audio data. The TTS
applies the coarticulation rules to the phoneme sequence and sends a
mouth animation sequence to the face modeler. The proposed THS can
make more natural lip sync and facial expression by using the face
animation generator than those using the conventional visemes only.
The experimental results show that our system has great potential for
the implementation of talking head for Korean text.
Abstract: This article investigates a contribution of synthesized visual speech. Synthesis of visual speech expressed by a computer consists in an animation in particular movements of lips. Visual speech is also necessary part of the non-manual component of a sign language. Appropriate methodology is proposed to determine the quality and the accuracy of synthesized visual speech. Proposed methodology is inspected on Czech speech. Hence, this article presents a procedure of recording of speech data in order to set a synthesis system as well as to evaluate synthesized speech. Furthermore, one option of the evaluation process is elaborated in the form of a perceptual test. This test procedure is verified on the measured data with two settings of the synthesis system. The results of the perceptual test are presented as a statistically significant increase of intelligibility evoked by real and synthesized visual speech. Now, the aim is to show one part of evaluation process which leads to more comprehensive evaluation of the sign speech synthesis system.
Abstract: We provide a supervised speech-independent voice recognition technique in this paper. In the feature extraction stage we propose a mel-cepstral based approach. Our feature vector classification method uses a special nonlinear metric, derived from the Hausdorff distance for sets, and a minimum mean distance classifier.
Abstract: This article presents a simple way to perform programmed voice commands for the interface with commercial Digital and Analogue Input/Output PCI cards, used in Robotics and Automation applications. Robots and Automation equipment can "listen" to voice commands and perform several different tasks, approaching to the human behavior, and improving the human- machine interfaces for the Automation Industry. Since most PCI Digital and Analogue Input/Output cards are sold with several DLLs included (for use with different programming languages), it is possible to add speech recognition capability, using a standard speech recognition engine, compatible with the programming languages used. It was created in this work a Visual Basic 6 (the world's most popular language) application, that listens to several voice commands, and is capable to communicate directly with several standard 128 Digital I/O PCI Cards, used to control complete Automation Systems, with up to (number of boards used) x 128 Sensors and/or Actuators.
Abstract: Mel Frequency Cepstral Coefficient (MFCC) features
are widely used as acoustic features for speech recognition as well
as speaker recognition. In MFCC feature representation, the Mel frequency
scale is used to get a high resolution in low frequency region,
and a low resolution in high frequency region. This kind of processing
is good for obtaining stable phonetic information, but not suitable
for speaker features that are located in high frequency regions. The
speaker individual information, which is non-uniformly distributed
in the high frequencies, is equally important for speaker recognition.
Based on this fact we proposed an admissible wavelet packet based
filter structure for speaker identification. Multiresolution capabilities
of wavelet packet transform are used to derive the new features.
The proposed scheme differs from previous wavelet based works,
mainly in designing the filter structure. Unlike others, the proposed
filter structure does not follow Mel scale. The closed-set speaker
identification experiments performed on the TIMIT database shows
improved identification performance compared to other commonly
used Mel scale based filter structures using wavelets.
Abstract: This paper investigates the performance of a speech
recognizer in an interactive voice response system for various coded
speech signals, coded by using a vector quantization technique namely
Multi Switched Split Vector Quantization Technique. The process of
recognizing the coded output can be used in Voice banking application.
The recognition technique used for the recognition of the coded speech
signals is the Hidden Markov Model technique. The spectral distortion
performance, computational complexity, and memory requirements of
Multi Switched Split Vector Quantization Technique and the
performance of the speech recognizer at various bit rates have been
computed. From results it is found that the speech recognizer is
showing better performance at 24 bits/frame and it is found that the
percentage of recognition is being varied from 100% to 93.33% for
various bit rates.
Abstract: Years of extensive research in the field of speech
processing for compression and recognition in the last five decades,
resulted in a severe competition among the various methods and
paradigms introduced. In this paper we include the different representations
of speech in the time-frequency and time-scale domains
for the purpose of compression and recognition. The examination of
these representations in a variety of related work is accomplished.
In particular, we emphasize methods related to Fourier analysis
paradigms and wavelet based ones along with the advantages and
disadvantages of both approaches.
Abstract: This work presents a novel means of extracting fixedlength parameters from voice signals, such that words can be recognized
in linear time. The power and the zero crossing rate are first
calculated segment by segment from a voice signal; by doing so, two
feature sequences are generated. We then construct an FIR system
across these two sequences. The parameters of this FIR system, used
as the input of a multilayer proceptron recognizer, can be derived by
recursive LSE (least-square estimation), implying that the complexity of overall process is linear to the signal size. In the second part of
this work, we introduce a weighting factor λ to emphasize recent
input; therefore, we can further recognize continuous speech signals.
Experiments employ the voice signals of numbers, from zero to nine, spoken in Mandarin Chinese. The proposed method is verified to
recognize voice signals efficiently and accurately.
Abstract: A state of the art Speaker Identification (SI) system requires a robust feature extraction unit followed by a speaker modeling scheme for generalized representation of these features. Over the years, Mel-Frequency Cepstral Coefficients (MFCC) modeled on the human auditory system has been used as a standard acoustic feature set for SI applications. However, due to the structure of its filter bank, it captures vocal tract characteristics more effectively in the lower frequency regions. This paper proposes a new set of features using a complementary filter bank structure which improves distinguishability of speaker specific cues present in the higher frequency zone. Unlike high level features that are difficult to extract, the proposed feature set involves little computational burden during the extraction process. When combined with MFCC via a parallel implementation of speaker models, the proposed feature set outperforms baseline MFCC significantly. This proposition is validated by experiments conducted on two different kinds of public databases namely YOHO (microphone speech) and POLYCOST (telephone speech) with Gaussian Mixture Models (GMM) as a Classifier for various model orders.
Abstract: A Web-based learning tool, the Learn IN Context
(LINC) system, designed and being used in some institution-s
courses in mixed-mode learning, is presented in this paper. This
mode combines face-to-face and distance approaches to education.
LINC can achieve both collaborative and competitive learning. In
order to provide both learners and tutors with a more natural way to
interact with e-learning applications, a conversational interface has
been included in LINC. Hence, the components and essential features
of LINC+, the voice enhanced version of LINC, are described. We
report evaluation experiments of LINC/LINC+ in a real use context
of a computer programming course taught at the Université de
Moncton (Canada). The findings show that when the learning
material is delivered in the form of a collaborative and voice-enabled
presentation, the majority of learners seem to be satisfied with this
new media, and confirm that it does not negatively affect their
cognitive load.
Abstract: Unlike general-purpose processors, digital signal
processors (DSP processors) are strongly application-dependent. To
meet the needs for diverse applications, a wide variety of DSP
processors based on different architectures ranging from the
traditional to VLIW have been introduced to the market over the
years. The functionality, performance, and cost of these processors
vary over a wide range. In order to select a processor that meets the
design criteria for an application, processor performance is usually
the major concern for digital signal processing (DSP) application
developers. Performance data are also essential for the designers of
DSP processors to improve their design. Consequently, several DSP
performance benchmarks have been proposed over the past decade or
so. However, none of these benchmarks seem to have included recent
new DSP applications.
In this paper, we use a new benchmark that we recently developed
to compare the performance of popular DSP processors from Texas
Instruments and StarCore. The new benchmark is based on the
Selectable Mode Vocoder (SMV), a speech-coding program from the
recent third generation (3G) wireless voice applications. All
benchmark kernels are compiled by the compilers of the respective
DSP processors and run on their simulators. Weighted arithmetic
mean of clock cycles and arithmetic mean of code size are used to
compare the performance of five DSP processors.
In addition, we studied how the performance of a processor is
affected by code structure, features of processor architecture and
optimization of compiler. The extensive experimental data gathered,
analyzed, and presented in this paper should be helpful for DSP
processor and compiler designers to meet their specific design goals.
Abstract: Independent component analysis (ICA) in the
frequency domain is used for solving the problem of blind source
separation (BSS). However, this method has some problems. For
example, a general ICA algorithm cannot determine the permutation
of signals which is important in the frequency domain ICA. In this
paper, we propose an approach to the solution for a permutation
problem. The idea is to effectively combine two conventional
approaches. This approach improves the signal separation
performance by exploiting features of the conventional approaches.
We show the simulation results using artificial data.
Abstract: The development of the signal compression
algorithms is having compressive progress. These algorithms are
continuously improved by new tools and aim to reduce, an average,
the number of bits necessary to the signal representation by means of
minimizing the reconstruction error. The following article proposes
the compression of Arabic speech signal by a hybrid method
combining the wavelet transform and the linear prediction. The
adopted approach rests, on one hand, on the original signal
decomposition by ways of analysis filters, which is followed by the
compression stage, and on the other hand, on the application of the
order 5, as well as, the compression signal coefficients. The aim of
this approach is the estimation of the predicted error, which will be
coded and transmitted. The decoding operation is then used to
reconstitute the original signal. Thus, the adequate choice of the
bench of filters is useful to the transform in necessary to increase the
compression rate and induce an impercevable distortion from an
auditive point of view.
Abstract: ICA which is generally used for blind source separation
problem has been tested for feature extraction in Speech recognition
system to replace the phoneme based approach of MFCC. Applying
the Cepstral coefficients generated to ICA as preprocessing has
developed a new signal processing approach. This gives much better
results against MFCC and ICA separately, both for word and speaker
recognition. The mixing matrix A is different before and after MFCC
as expected. As Mel is a nonlinear scale. However, cepstrals
generated from Linear Predictive Coefficient being independent
prove to be the right candidate for ICA. Matlab is the tool used for
all comparisons. The database used is samples of ISOLET.