Plosive/fricative distinction: the voiceless case

Area of Science:

• Speech processing
• Acoustic phonetics
• Computational linguistics

Background:

• Distinguishing between voiceless plosives and fricatives is crucial for speech recognition and linguistic analysis.
• Previous methods often require complex acoustic features or extensive training data.

Purpose of the Study:

• To develop a computationally efficient algorithm for classifying voiceless plosives and fricatives.
• To evaluate the effectiveness of using zero-crossing rate, log rms energy, and rate of rise (ROR) for this classification task.

Main Methods:

• Utilized three waveform measures: zero-crossing rate, logarithm of root-mean-square (rms) energy, and rate of rise (ROR).
• Identified significant peaks in the ROR contour, using log rms energy and zero-crossing rate for validation.
• Employed the magnitude of the first significant ROR peak as the primary classification feature.
• Trained and tested the algorithm on 1364 speech tokens from multiple speakers and positions.

Main Results:

• Achieved an overall correct classification rate of 96.8% for distinguishing voiceless plosives from fricatives.
• Demonstrated the robustness of the algorithm across word-initial, medial, and final positions.
• Validated the effectiveness of the selected acoustic features for phoneme discrimination.

Conclusions:

• The proposed method offers a highly accurate and efficient approach for plosive/fricative distinction.
• This technique has significant implications for improving automatic speech recognition systems.
• The findings suggest that simple waveform analysis can yield powerful results in phonetics.