Signal Processing and Speech Analysis 2
HomeFull List of Titles
1: ICSLP'98 Proceedings
Keynote SpeechesText-To-Speech Synthesis 1Spoken Language Models and Dialog 1Prosody and Emotion 1Hidden Markov Model Techniques 1Speaker and Language Recognition 1Multimodal Spoken Language Processing 1Isolated Word RecognitionRobust Speech Processing in Adverse Environments 1Spoken Language Models and Dialog 2Articulatory Modelling 1Talking to Infants, Pets and LoversRobust Speech Processing in Adverse Environments 2Spoken Language Models and Dialog 3Speech Coding 1Articulatory Modelling 2Prosody and Emotion 2Neural Networks, Fuzzy and Evolutionary Methods 1Utterance Verification and Word Spotting 1 / Speaker Adaptation 1Text-To-Speech Synthesis 2Spoken Language Models and Dialog 4Human Speech Perception 1Robust Speech Processing in Adverse Environments 3Speech and Hearing Disorders 1Prosody and Emotion 3Spoken Language Understanding Systems 1Signal Processing and Speech Analysis 1Spoken Language Generation and Translation 1Spoken Language Models and Dialog 5Segmentation, Labelling and Speech Corpora 1Multimodal Spoken Language Processing 2Prosody and Emotion 4Neural Networks, Fuzzy and Evolutionary Methods 2Large Vocabulary Continuous Speech Recognition 1Speaker and Language Recognition 2Signal Processing and Speech Analysis 2Prosody and Emotion 5Robust Speech Processing in Adverse Environments 4Segmentation, Labelling and Speech Corpora 2Speech Technology Applications and Human-Machine Interface 1Large Vocabulary Continuous Speech Recognition 2Text-To-Speech Synthesis 3Language Acquisition 1Acoustic Phonetics 1Speaker Adaptation 2Speech Coding 2Hidden Markov Model Techniques 2Multilingual Perception and Recognition 1Large Vocabulary Continuous Speech Recognition 3Articulatory Modelling 3Language Acquisition 2Speaker and Language Recognition 3Text-To-Speech Synthesis 4Spoken Language Understanding Systems 4Human Speech Perception 2Large Vocabulary Continuous Speech Recognition 4Spoken Language Understanding Systems 2Signal Processing and Speech Analysis 3Human Speech Perception 3Speaker Adaptation 3Spoken Language Understanding Systems 3Multimodal Spoken Language Processing 3Acoustic Phonetics 2Large Vocabulary Continuous Speech Recognition 5Speech Coding 3Language Acquisition 3 / Multilingual Perception and Recognition 2Segmentation, Labelling and Speech Corpora 3Text-To-Speech Synthesis 5Spoken Language Generation and Translation 2Human Speech Perception 4Robust Speech Processing in Adverse Environments 5Text-To-Speech Synthesis 6Speech Technology Applications and Human-Machine Interface 2Prosody and Emotion 6Hidden Markov Model Techniques 3Speech and Hearing Disorders 2 / Speech Processing for the Speech and Hearing Impaired 1Human Speech ProductionSegmentation, Labelling and Speech Corpora 4Speaker and Language Recognition 4Speech Technology Applications and Human-Machine Interface 3Utterance Verification and Word Spotting 2Large Vocabulary Continuous Speech Recognition 6Neural Networks, Fuzzy and Evolutionary Methods 3Speech Processing for the Speech-Impaired and Hearing-Impaired 2Prosody and Emotion 72: SST Student Day
SST Student Day - Poster Session 1SST Student Day - Poster Session 2Author Index
A B C D E F G H I
J K L M N O P Q R
S T U V W X Y Z
Multimedia Files
Improving Pitch Estimation with Short Duration Speech Samples
Authors:
William A. Ainsworth, Department of Communication & Neuroscience, Keele University (U.K.)
Charles R. Day, Department of Communication & Neuroscience, Keele University (U.K.)
Georg F. Meyer, Department of Communication & Neuroscience, Keele University (U.K.)
Page (NA) Paper number 512
Abstract:
Hermes' Sub Harmonic Summation (SHS) pitch determination algorithm is an effective technique for extracting the percept of pitch from human speech. Effective determination of the pitch in a passage of speech is believed to be fundamental for higher level speech processing applications such as speech or speaker recognition. Of particular interest is the need to extract pitch from speech in less than ideal conditions eg. in the presence of noise or using very short analysis windows. In an attempt to deliver accurate pitch estimates from relatively short analysis windows this paper describes an evaluation of two forms of the SHS procedure: in one case, FFT-SHS, the procedure uses the conventional Fast Fourier Transform (FFT) in its spectral analysis step; in the second case, RAFT-SHS, the ReAssigned Fourier Transform (RAFT) technique is used instead of the FFT.
SL980512.PDF (From Author)
SL980512.PDF (Rasterized)
TOP
An Instantaneous-Frequency-Based Pitch Extraction Method for High-Quality Speech Transformation: Revised TEMPO in the STRAIGHT-Suite
Authors:
Hideki Kawahara, Wakayama University/ATR/CREST (Japan)
Alain de Cheveigné, Paris 7 University/CNRS (France)
Roy D. Patterson, CNBH University of Cambridge (U.K.)
Page (NA) Paper number 659
Abstract:
A new source information extraction algorithm is proposed to provide a reliable source signal for an extremely high-quality speech analysis, modification, and transformation system called STRAIGHT-suite (Speech Transformation and Representation based on Adaptive Interpolation of weiGHTed spectrogram). The proposed method makes use of instantaneous frequencies in harmonic components based on their reliability. A performance evaluation is conducted using a simultaneous EGG (Electroglottograph) recording as the reference signal. The error variance for F0 extraction using the proposed algorithm is shown to be about 1/3 that of the previous F0 extraction method used in STRAIGHT-suite, although the previous algorithm is still competitive with conventional F0 extraction methods.
SL980659.PDF (From Author)
SL980659.PDF (Rasterized)
TOP
Speaker-Independent Speech Recognition Using Micro Segment Spectrum Integration
Authors:
Kiyoaki Aikawa, NTT Human Interface Laboratories, Speech and Acoustics Laboratories (Japan)
Page (NA) Paper number 262
Abstract:
This paper proposes a new spectral estimation method for automatic speech recognition. The spectrum estimated with the conventional data window of around 30 ms shows harmonic structure in the voiced portions of speech data. The harmonic frequency interval is often comparable to the formant frequency interval for female voices with high F0, which results in spectral estimation error. The new idea is to estimate spectrum by taking the Lp norm of the time series of the spectrum obtained from a very short speech segment. The new method, called the micro-segment spectrum integration, provides (1) precise spectral estimation not affected by harmonic structure, and (2) noise-robustness by suppressing noisy speech segments. Phoneme recognition experiments demonstrate that the micro-segment spectrum integration method outperforms conventional spectral estimation methods.
SL980262.PDF (From Author)
SL980262.PDF (Rasterized)
TOP
On Robust Speech Analysis Based On Time-Varying Complex AR Model
Authors:
Keiichi Funaki, Hokkaido University (Japan)
Yoshikazu Miyanaga, Hokkaido University (Japan)
Koji Tochinai, Hokkaido University (Japan)
Page (NA) Paper number 1001
Abstract:
We have already developed time-varying complex AR (TV-CAR) parameter estimation based on minimizing mean square error (MMSE) for analytic speech signal. Although the MMSE approach is commonly and successfully applied in various parameter estimation such as conventional LPC, it is well-known that an MMSE method easily suffers from biased and inaccurate spectrum estimation due to non-Gaussian nature of glottal excitation for voiced speech in the context of speech analysis. This paper offers robust parameter estimation algorithm for the TV-CAR model by applying Huber's robust M-estimation approach and two kinds of robust algorithms are derived: Newton-type algorithm and weighted least squares (WLS) algorithm. The preliminary experiments with synthetic signal generated by glottal source model excitation and natural speech uttered by female speaker demonstrate that the time-varying complex AR method is sufficiently robust against non-Gaussian nature of glottal source excitation owing to the improved resolution in the frequency domain.
SL981001.PDF (Scanned)
TOP
Spectral Basis Functions from Discriminant Analysis
Authors:
Hynek Hermansky, Oregon Graduate Institute Of Science And Technology (USA)
Narendranath Malayath, Oregon Graduate Institute Of Science And Technology (USA)
Page (NA) Paper number 616
Abstract:
The work examines Karhunen-Loeve Transform and Linear Discriminant Analysis as means for designing optimized spectral bases for the projection of the critical-band auditory-like spectrum.