Speech Pause Distribution Analysis for Cognitive Impairment Detection

The patent segments the speech sample into multiple pause segments and calculates pause duration for each segment. It then determines the distribution of pause durations and calculates the distance between distributions across multiple tasks, rather than using a single mean pause duration. This segmentation approach allows the system to capture subtle changes and inter-individual variability while maintaining computational feasibility.

The patent transitions from using a single scalar feature (mean pause duration) to using distribution-based features across multiple tasks. By calculating the distance between distributions of pause durations from different tasks, the system adds a dimensional aspect that captures inter-individual variability and subtle cognitive changes, thereby improving detection precision without excessive complexity.

The patent divides the speech sample into discrete pause segments and systematically processes each segment to determine pause duration. This segmentation allows the complex task of analyzing multiple speech samples to be broken down into manageable steps, improving discrimination accuracy while keeping processing complexity controlled through structured analysis.

The patent changes the parameter from simple mean pause duration to distribution-based pause duration metrics. By calculating statistical parameters (mean, standard deviation) of pause durations across multiple tasks and computing distances between these distributions, the system enhances discrimination accuracy. The parameter transformation is systematic and can be implemented efficiently, balancing improved detection with manageable processing complexity.