Recurrence Plot Speech Segmentation Boundary Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing speech segmentation methods fail to provide a statistical-based approach for detecting the boundary of coarticulated units from isolated speech using recurrence plots, particularly in non-linear systems, leading to ineffective processing of non-linear speech signals and false alarms.
Innovation Solution
A method and system that utilize recurrence plots to detect coarticulated unit boundaries by storing speech signals in vectors, defining time windows, calculating recurrence matrices, shifting non-overlapping windows, finding similarities based on distance measures, and generating similarity scores to flag coarticulated boundaries, enabling determinism in non-linear speech processing for automatic speech segmentation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If supervised learning techniques are used for speech segmentation, then classification accuracy can be achieved, but false alarming increases and reliability decreases
Solution Approach 1:
The patent replaces supervised learning techniques (mechanical/classical approach) with recurrence plot analysis (non-linear dynamics approach). This substitution eliminates the need for training sets and classifiers, thereby reducing false alarms while maintaining segmentation accuracy through intrinsic system dynamics analysis.
Solution Approach 2:
The recurrence plot method allows the speech signal to reveal its own structure and boundaries through self-comparison in phase space. The system serves itself by detecting coarticulated unit boundaries through intrinsic recurrence patterns without external classification, reducing false alarming caused by supervised learning limitations.
2Loss of information
If conventional recurrence plot methods are used for speech analysis, then visual patterns can be obtained, but statistical-based detection of coarticulated unit boundaries is not achieved
Solution Approach 1:
The patent transforms the recurrence plot from a purely visual tool to a statistical measurement tool by introducing distance measures between successive recurrence matrices. This parameter change enables quantitative boundary detection while preserving the visual pattern information, achieving both objectives simultaneously.
Solution Approach 2:
The patent adds a temporal dimension to the recurrence plot analysis by comparing successive recurrence matrices over time. This dimensional extension transforms static visual patterns into dynamic statistical measurements, enabling precise detection of coarticulated unit boundaries while retaining visual pattern information.
Data Source
AI summary
The application provides a method and system for determinism in non-linear systems for speech processing, particularly automatic speech segmentation for building speech recognition systems. More particularly, the application enables a method and system for detecting boundary of coarticulated units from isolated speech using recurrence plot.