Fluency Profiling System: an automated system for analyzing the temporal properties of speech

Area of Science:

• Linguistics
• Cognitive Science
• Speech Processing

Background:

• Analyzing speech and pause durations reveals temporal characteristics.
• Natural speech analysis is hindered by noise and difficulty classifying pause types (articulation vs. cognitive).
• Existing methods often discard valuable cognitive information present in natural speech pauses.

Purpose of the Study:

• To develop a fully automated system for analyzing temporal parameters in natural speech.
• To accurately determine speech-pause transitions and estimate duration distributions.
• To overcome limitations of previous methods in handling noise and classifying pauses.

Main Methods:

• Utilized Gaussian mixture models (GMMs) for data distribution analysis.
• Applied GMMs to identify theoretical components within speech and pause data.
• Developed automated classification of speech components and duration computation.

Main Results:

• Successfully automated the identification of speech-pause transitions.
• Enabled accurate estimation of temporal parameters for both speech and pause durations.
• Provided a robust method for classifying pause segments, distinguishing between articulatory and cognitive pauses.

Conclusions:

• The automated system effectively analyzes temporal characteristics of natural speech.
• Gaussian mixture models are crucial for accurate speech-pause segmentation and duration analysis.
• This approach enhances the study of cognitive processes through natural speech analysis.