Mamoru IWAKI
Associate Professor,
Dr. Eng.
NOTE: Japanese page is here.
OUTLINE
- Research Fields
- Speech & Audio Signal Processing, Digital Signal Processing, Approximation & Interpolation Theory
- Research Interests (details)
- I am interested in theory and implementation of waveform analysis
for speech and audio signal processing.
My research currently focuses on an instantaneous analysis of waveforms by
the natural observation theory and its implementation by digital filters.
- Publications (details)
-
Papers & Letters
Proceedings of Intl. Conf.
Tech. Reports & Others
- Others (details)
-
Last modified on Dec. 24, 1996 by
Research Interests
- 1. Fundamental Research on Analysis and Recognition of Waveforms
-
Instantaneous analysis methods of waveforms are investigated based on the
Natural Observation Theory proposed by Prof. Iijima.
The theory was constructed in the aspect of waveform observation:
a waveform can be observed from moment to moment as time goes by.
The concept of natural observation method was exported to discrete-time
waveforms, and it was implemented on digital computers.
Digital filters for synthesis and recognition of waveforms are developed
and applied to speech and audio signal processing.
- 2. Analysis of Individuality in Glottal Vibration
-
Speech waveforms contain not only phonetic information but also
prosody which can express individuality, emotion and so on.
In particular, individuality is essential for advanced speech processing in
the years ahead.
In conventional investigations of individualities, it has become clear
that pitch pattern and higher band in spectral envelopes contain an
important information of individuality, however, it is still hard to
know how the glottal vibration has effects on it.
Hence we develop an effective pitch extraction method in order to make the
aspect of glottal vibration clear.
EGG waveforms from Laryngograph are used as both reference and target
signal for analysis.
- 3. Representation Method of Signals for Digital Signal Processing
-
Sampling is indispensable for digital signal processing of continuous-time
signals.
It is hard to realize effective processing without appropriate sampling.
Considering all stages of signal processing, i.e. generation, observation,
sampling, processing, and output, we discuss a sampling method for
effective digital signal processing and its applications.
- 4. Design Methods and Applications of Digital Filter
-
Digital filter design method is one of the solutions of optimization
problems: the desired impulse response/transfer function is obtained as an
approximation function based on a certain optimal criteria, e.g. minimax
criteria.
Numerical analysis and improvement of Remez-exchange method, which is
known as a solution for minimax criteria, are investigated.
And simultaneous approximation of amplitude and phase characteristics is
also investigated for Linear phase FIR digital filters under the criteria.
Back to the OUTLINE
Selections from the Latest Publications
- Pluralizing Method of Simple Similarity Based on Natural Observation Method of Normal Type --- with an application to Speaker-Independent Vowel Recognition ---,
Trans. IEICE (A), 1997 (In Japanese)(Printing).
- On Uncertainty Principle in Natural Observation Transform
of Normal Type,
Trans. IEICE (A), 1997 (In Japanese)(Printing).
- On Uncertainty Principle in Natural Observation Transform
of Neighboring Type,
Trans. IEICE (A), Vol.J79-A, No.11, 1996 (In Japanese).
- Consideration on Derivation Principles of Natural Observation Method of Discrete type, Trans. IEICE (A), Vol.J79-A, No.11, 1996 (In Japanese).
- Description of FIR digital filters in the form of parallel
connection of linear phase FIRs,
Electronics Letters, Vol.32, No.11, 1996.
- Natural observation method for discrete-time waveforms,
Trans. IEICE (A), Vol.J79-A, No.3, 1996 (In Japanese).
- On a Vowel Synthesis System using Natural Observation Method,
Tech. Rep. IEICE, SP95-142, 1996 (In Japanese).
- Pluralizing Method of Similarity for Speaker-Independent Vowel
Recognition,
Tech. Rep. IEICE, SP95-141, 1996 (In Japanese).
- Fundamental theory of natural observation method with complete
reconstruction property by finite sum
--- Natural observation theory of normal type ---,
Trans. IEICE (A), Vol.J79-A, No.1, 1996 (In Japanese).
- Structural analysis for vowel waveforms by using finite
Dirichlet series,
Tech. Rep. Hear. Acoust. Soc. Jpn., H-94-42, 1994 (In Japanese).
- Sampling theorem for spline signal space of arbitrary degree,
IEICE Trans. Fundamentals, Vol.E77-A, No.5, 1994.
- Polynomial interpolation for Remez exchange method,
Electronics Letters, Vol.28, No.20, 1992.
- Periodic sampling basis and its biorthonormal basis
for the signal spaces of piecewise polynomials,
Trans. IEICE (A), Vol.J75-A, No.6, 1992 (In Japanese).
Others.
Back to the OUTLINE
EOF