Speech Duration Detector Trailing End Noise Resolution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing speech-duration detection methods, such as those using finite state automata, face challenges in accurately detecting the trailing end of speech due to noise interference, leading to incorrect detection and failure to distinguish between silence in words and post-utterance silence.
Innovation Solution
A speech-duration detector employing a characteristic extracting unit, starting-end detecting unit, trailing-end-candidate detecting unit, and trailing-end-candidate determining unit, which utilize multiple time length parameters to differentiate between speech and noise, ensuring accurate detection of speech starts and ends by using a finite state automaton with specific state transitions based on power threshold comparisons.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the trailing end detection time is reduced to avoid noise interference, then noise resistance is improved, but speech duration detection precision deteriorates because words with double consonants are incorrectly divided
Solution Approach 1:
The detection process is segmented into two distinct phases: a first trailing end detection process that operates during speech to detect candidate points, and a second trailing end detection process that operates after speech ends to confirm the actual trailing end. This segmentation allows the system to use different detection criteria appropriate for each phase, resolving the contradiction between noise resistance and precision.
Solution Approach 2:
The first trailing end detection process performs preliminary detection of candidate trailing end points during the speech period. These candidate points are then validated in the second detection process after speech ends. This preliminary action allows the system to prepare potential trailing end candidates without committing to a final detection, thereby avoiding premature termination due to noise while maintaining accuracy.
2Device complexity
If a single time control parameter is used for trailing end detection, then device complexity is reduced, but detection accuracy deteriorates due to inability to distinguish word silence from post-utterance silence
Solution Approach 1:
The detection system dynamically switches between two detection modes based on the speech state. During speech (first detection process), the system detects candidate trailing end points using one set of criteria. After speech ends (second detection process), it switches to confirmation mode using different criteria. This dynamic adaptation allows accurate distinction between word silences and post-utterance silences without requiring excessive parameters.
Solution Approach 2:
The detection operates in periodic cycles: the first trailing end detection process runs during speech to identify candidates, then pauses, and the second detection process runs after speech to confirm the trailing end. This periodic alternation between detection and confirmation phases enables accurate trailing end detection while maintaining relatively simple device structure.
Data Source
AI summary
A speech-duration detector includes a starting-end detecting unit that detects a starting end of a first duration where the characteristic exceeds a threshold value as a starting end of a speech-duration, when the first duration continues for a first time length; a trailing-end-candidate detecting unit that detects a starting end of a second duration where the characteristic is lower than the threshold value as a candidate point for a trailing end of speech, when the second duration continues for a second time length; and a trailing-end-candidate determining unit that determines the candidate point as a trailing end of the speech-duration, when the second duration where the characteristic exceeds the threshold value does not continue for the first time length while a third time length elapses from measurement at the candidate point.