Related Experiment Video
Updated: Aug 9, 2026
09:09
Foreign Accent and Forensic Speaker Identification in Voice Lineups: The Influence of Acoustic Features Based on Prosody
Published on: September 27, 2024
Evaluation of a spectrum target prediction model in speech perception
1ATR Auditory and Visual Perception Research Laboratories, Kyoto, Japan.
The Journal of the Acoustical Society of America
|February 1, 1990
Summary
This study introduces a novel spectrum target prediction model that accurately estimates spectral targets from short speech sequences. The model effectively reduces transitional sounds and recovers coarticulated vowel characteristics for speech processing applications.
Area of Science:
- Speech Signal Processing
- Acoustics
- Computational Linguistics
Background:
- Coarticulation neutralizes vowel characteristics in speech.
- Accurate spectral target prediction is crucial for speech processing.
- Existing models struggle with short-period spectrum sequences and identifying spectral transition onsets.
Purpose of the Study:
- To propose and evaluate a novel spectrum target prediction mechanism.
- To estimate target values from short-period spectrum sequences without explicit onset detection.
- To recover vowel characteristics affected by coarticulation and improve syllable transition stability.
Main Methods:
- Approximating the cepstrally smoothed LPC spectrum trajectory using a second-order critically damped system.
- Utilizing short-period spectrum sequences (50 ms) for prediction.
- Comparing model predictions with psychoacoustic experimental results.
Main Results:
- The model accurately estimates target values from short-period spectrum sequences.
- It successfully reduces the length of transitional sounds.
- Vowel characteristics neutralized by coarticulation are recovered, and stable syllable transition characteristics are extracted.
Conclusions:
- The proposed model offers a robust method for spectrum target prediction in speech processing.
- It effectively handles coarticulation and improves the analysis of transitional speech sounds.
- Applicable for coarticulation recovery and enhancing speech signal processing algorithms.
