Dual Voice Recognition Module for Mixed-Language Input

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current speech recognition systems in vehicles face challenges in accurately processing voice inputs across languages, especially when users are not familiar with the local language, due to limited computing resources and complex pronunciation rules, leading to inefficiencies in embedded systems.

Innovation Solution

A dual speech recognition module system that processes voice inputs differently based on user language and auxiliary languages, allowing for efficient recognition by subdividing inputs into sections and using distinct rule systems, with one module handling navigation commands in the user language and another handling address data in auxiliary languages, thereby reducing computational load and improving accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If speech recognition systems process all speech segments using a single comprehensive rule set covering multiple languages, then recognition accuracy across languages improves, but computing time and processing load increase significantly

Engineering Contradiction:
Improvespeech recognition accuracyVSAvoidcomputing time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The speech recognition system divides the speech input into multiple speech segments and processes each segment with appropriate language rules. The system segments speech based on language detection, allowing different portions of the same speech input to be processed by different language-specific rule sets, thereby reducing overall processing time while maintaining multi-language accuracy.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

Different language rules are applied to different speech segments based on the detected language of each segment. Instead of applying a single comprehensive rule set to all speech, the system uses language-appropriate rules locally for each segment, improving recognition accuracy for that segment while avoiding the computational overhead of processing all segments with all possible language rules.

Inventive Principle:
Principle #3Local quality

2Adaptability or versatility

If a single speech recognition module uses comprehensive language rules for all languages, then recognition capability across languages improves, but device complexity increases

Engineering Contradiction:
Improvemulti-language recognition capabilityVSAvoidspeech recognition system complexity
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The speech recognition system achieves multi-language capability through a universal architecture that can dynamically select and apply different language-specific rule sets. Rather than building separate specialized modules for each language, a single speech recognition module is designed to handle multiple languages by selecting appropriate rules based on language detection, reducing overall system complexity.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Solution Approach 2:

The system changes the operational parameters of the speech recognition module based on the detected language of each speech segment. By dynamically adjusting which language rules are applied based on language identification, the system maintains versatility across languages without requiring permanent, complex configurations for every possible language scenario.

Inventive Principle:
Principle #35Parameter changes

3Productivity

If speech recognition processes entire speech inputs in one language rule set, then processing speed improves, but recognition accuracy decreases for mixed-language inputs

Engineering Contradiction:
Improvespeech processing speedVSAvoidspeech recognition accuracy
Core Design Contradiction:
ProductivityVSMeasurement precision

Solution Approach 1:

The system segments speech inputs into language-specific portions and processes each segment with the appropriate language rules. This segmentation allows the system to maintain fast processing by avoiding comprehensive multi-language analysis for every segment, while simultaneously preserving accuracy by applying the correct language-specific rules to each segment.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The system performs preliminary language detection and speech segmentation before applying language-specific recognition rules. By identifying the language of each speech segment in advance, the system can quickly apply the appropriate rules without needing to reconsider language possibilities during processing, thus maintaining speed while ensuring accuracy.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP3152753B1Assistance system that can be controlled by means of voice inputs, having a functional device and a plurality of voice recognition modules
Publication Date: 2019.03.13 CONTINENTAL AUTOMOTIVE GMBH
  • EP3152753B1 patent drawingFigure 1~2
  • EP3152753B1 patent drawingFigure 3~4
  • EP3152753B1 patent drawingFigure 5

AI summary

The invention relates to an assistance system (3) that can be controlled at least to some extent by means of voice inputs (12), having a functional device (6) and a voice recognition device (5), wherein the voice recognition device has at least one first voice recognition module (9), which attributes voice inputs (12) a meaning on the basis of a first system of rules, and a second voice recognition module (10), which attributes voice inputs (12) a meaning on the basis of a second system of rules, wherein the results from one of the two voice recognition modules (9, 10) are respectively taken as a basis for ascertaining which voice sections (12a, 12b) are processed by the first voice recognition module (9) and which voice sections are processed by the second voice recognition module (10). A first voice recognition module (9) can therefore recognise a work command, for example a navigation command, while a second voice recognition module (10) can recognise, in an auxiliary language, an address statement that is not stored owing to unavailable translations of address databases into the user language.