Adaptive Silence Detection in Speech Recognition Systems

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current speech recognition systems face challenges in accurately differentiating between natural pauses and the end of speech, particularly in real-time systems, due to limitations in silence detection methods.

Innovation Solution

A computer-implemented method for speech recognition that adjusts silence detection parameters based on lead and terminal periods, threshold amplitude, and environmental factors, using adaptive techniques such as statistical estimation and pulse code modulation, to accurately terminate speech recognition.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If fixed silence detection parameters are used, then the system is simple to operate, but the accuracy of differentiating between pauses and end of speech deteriorates

Engineering Contradiction:
Improvesilence detection accuracyVSAvoidparameter adjustment complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent implements dynamic adjustment of silence detection parameters (lead period, terminal period, threshold amplitude) based on real-time analysis of audio signal characteristics. The system continuously adapts these parameters during operation to match the specific speech patterns and environmental conditions, transforming fixed parameters into dynamic, context-aware values that improve detection accuracy without requiring complex manual configuration

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The system performs self-adjustment of silence detection parameters by automatically analyzing the audio signal characteristics and optimizing parameter values without external intervention. The algorithm independently determines optimal lead periods, terminal periods, and threshold amplitudes based on the observed speech patterns, enabling the system to adapt to different users and environments autonomously

Inventive Principle:
Principle #25Self-service

2Measurement precision

If adaptive silence detection is implemented, then the accuracy of speech recognition is improved, but the computational resources required increase

Engineering Contradiction:
Improvespeech recognition accuracyVSAvoidcomputational energy consumption
Core Design Contradiction:
Measurement precisionVSUse of energy by moving object

Solution Approach 1:

The patent applies partial adaptation by adjusting only the critical silence detection parameters (lead period, terminal period, threshold amplitude) based on essential signal characteristics rather than performing complete signal processing analysis. This selective adaptation achieves sufficient accuracy improvement while limiting computational overhead to only what is necessary for effective silence detection

Inventive Principle:
Principle #16Partial or excessive action

Solution Approach 2:

The system optimizes computational efficiency by changing specific parameters (lead period, terminal period, threshold amplitude) rather than redesigning the entire speech recognition system. By focusing parameter adjustments on the silence detection module only, the patent achieves accuracy improvement with minimal additional computational burden

Inventive Principle:
Principle #35Parameter changes

3Adaptability or versatility

If silence detection parameters are adjusted frequently, then the adaptability to different speech patterns is improved, but the system stability deteriorates

Engineering Contradiction:
Improveadaptability to speech patternsVSAvoidsystem stability
Core Design Contradiction:
Adaptability or versatilityVSStability of the object's composition

Solution Approach 1:

The patent implements periodic adjustment of silence detection parameters rather than continuous adjustment. The system evaluates speech patterns over defined periods and adjusts parameters at appropriate intervals, allowing the system to adapt to different speech patterns while maintaining stability during each evaluation period. This periodic approach prevents excessive parameter changes that could destabilize the system

Inventive Principle:
Principle #19Periodic action

Data Source

PatentEP4264598B1A computer implemented method and an apparatus for silence detection in speech recognition
Publication Date: 2024.09.25 ELISA OYJ
  • EP4264598B1 patent drawingFigure 1~2
  • EP4264598B1 patent drawingFigure 3
  • EP4264598B1 patent drawingFigure 4A~4C

AI summary

A computer implemented method for speech recognition from an audio signal is provided.The method comprises: obtaining initial values for silence detection parameters comprising:a lead period; a threshold amplitude; and a terminal period; starting to detect an amplitude of the audio signal at a first time T1 of the audio signal; optionally, adjusting the threshold amplitude based on the detected amplitude; starting the speech recognition from a second time T2 of the audio signal, wherein the second time T2 is at or after the first time T1; starting silence detection from the audio signal when lead period has elapsed after the second time T2 comprising: responsive to detecting amplitude below the threshold amplitude for a duration of the terminal period, terminating the speech recognition and the silence detection at a third time T3 of the audio signal; and adjusting the silence detection parameters based on the detected amplitude changes of the audio signal between the first time T1 and the third time T3.