Dynamic Threshold Speech Recognition Flexibility

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing speech recognition systems face challenges in flexibility due to fixed threshold settings, which are inadequate for varying user environments and device availability, leading to suboptimal processing results in different operation modes.

Innovation Solution

A speech processing apparatus and method that dynamically sets a threshold for determining the validity of speech recognition results based on operation modes, comparing the certainty factor of recognition results with mode-specific thresholds to effectively utilize local or server-based recognition.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If a fixed threshold is used for determining speech recognition results, then the processing method is simple, but the flexibility is insufficient for varying user environments and device availability

Engineering Contradiction:
Improveflexibility of processingVSAvoidcomplexity of threshold management
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent implements dynamic threshold adjustment by setting different threshold values based on operation modes (local recognition mode, server recognition mode, hybrid mode). The threshold is no longer fixed but adapts to the current operational context, allowing the system to flexibly respond to varying user environments and device availability while maintaining manageable complexity through mode-based classification.

Inventive Principle:
Principle #15Dynamics

2Measurement precision

If speech recognition is performed only by a local terminal, then the response speed is fast, but the recognition accuracy may be insufficient in complex environments

Engineering Contradiction:
Improverecognition accuracyVSAvoidresponse time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent segments the speech recognition function into two parts: local recognition (fast response) and server recognition (high accuracy). The system performs initial recognition locally and then selectively sends results to the server for verification or re-recognition based on certainty factors and operation modes. This segmentation allows the system to achieve both fast response and high accuracy by combining the strengths of both approaches.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent introduces an intermediary mechanism (the determination unit and operation mode controller) that mediates between local and server recognition. This intermediary evaluates recognition results using certainty factors and operation modes to decide whether to accept local results or request server verification, thereby optimizing the balance between response speed and recognition accuracy without requiring all recognition to go through the server.

Inventive Principle:
Principle #24Intermediary (Mediator)

3Measurement precision

If speech recognition is always performed by both local terminal and server, then the recognition accuracy is improved, but the processing load and time consumption increase

Engineering Contradiction:
Improverecognition accuracyVSAvoidprocessing efficiency
Core Design Contradiction:
Measurement precisionVSProductivity

Solution Approach 1:

The patent applies partial action by performing server recognition only when necessary (based on operation modes and certainty factors) rather than always. In high-certainty scenarios or when operation modes indicate local sufficiency, only local recognition is performed. This partial approach maintains high recognition accuracy when needed while improving processing efficiency by avoiding unnecessary server communications and computations.

Inventive Principle:
Principle #16Partial or excessive action

Solution Approach 2:

The patent changes the parameter of recognition configuration based on operation modes and certainty factors. Instead of using a fixed dual-recognition approach, the system dynamically adjusts whether to use local-only, server-only, or hybrid recognition by changing the operational parameters (thresholds, mode selections). This parameter-based flexibility optimizes the balance between recognition accuracy and processing efficiency for different scenarios.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP3428917B1Voice processing device and voice processing method
Publication Date: 2021.05.19 SONY GROUP CORP
  • EP3428917B1 patent drawingFigure 1
  • EP3428917B1 patent drawingFigure 2
  • EP3428917B1 patent drawingFigure 3

AI summary

[Object] To improve the flexibility of processing regarding speech recognition. [Solution] A speech processing apparatus including: a determination unit configured to determine how to deal with a recognition result of speech data obtained by a first speech recognizer on a basis of a comparison between a certainty factor of the recognition result of the speech data obtained by the first speech recognizer and a threshold; and a threshold setting unit configured to set dynamically the threshold.