Vehicle Speech Recognition Using Priority-Based Signal Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional wake-up voice-based speech recognition in vehicles faces challenges due to environmental noise, misrecognition of keywords, and incorrect activation in environments like multimedia playback and vehicle noise, leading to decreased performance.
Innovation Solution
A method and device that collect and analyze various types of information within the vehicle, such as speech, vehicle, image, and sensing information, to adaptively prioritize and process signals for accurate wake-up voice and command voice recognition, using multiple microphones and adjusting their reliability based on occupancy and environmental conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If conventional wake-up voice-based speech recognition is used in vehicles, then the system can activate speech recognition function through voice input, but the system frequently wakes up incorrectly due to environmental noise, multimedia playback, and vehicle noise
Solution Approach 1:
The patent segments the speech recognition system into multiple independent components: wake-up word recognition module, keyword recognition module, and peripheral device information collection modules. Each module processes specific aspects of the input signal independently, allowing the system to distinguish between actual wake-up commands and environmental noise more effectively.
Solution Approach 2:
The patent introduces peripheral devices (microphones, cameras, sensors) as intermediaries that collect environmental information and provide it to the speech recognition system. These intermediaries help filter and contextualize the voice input, enabling the system to distinguish between genuine wake-up commands and spurious signals from multimedia playback or vehicle noise.
2Productivity
If the system activates speech recognition frequently due to incorrect wake-up recognition, then the system responds to perceived commands, but the speech recognition performance deteriorates due to continuous activation
Solution Approach 1:
The patent performs preliminary analysis of the wake-up word using multiple microphones and peripheral devices before fully activating the speech recognition system. By pre-processing and validating the wake-up signal in advance, the system avoids unnecessary full activations and maintains higher recognition accuracy when speech recognition is actually needed.
Solution Approach 2:
The patent implements feedback mechanisms where the system continuously monitors environmental noise levels, multimedia playback status, and recognition confidence scores. This feedback allows the system to adjust its wake-up sensitivity dynamically, reducing false activations while maintaining responsiveness to legitimate commands.
3Measurement precision
If multiple microphones and peripheral devices are used for speech recognition, then the accuracy of wake-up voice recognition improves, but the device complexity increases
Solution Approach 1:
The patent designs peripheral devices to serve multiple functions: microphones are used for both wake-up detection and keyword recognition, cameras provide both visual context and attention detection, and sensors serve both environmental monitoring and noise cancellation. This multi-functionality reduces the need for dedicated components for each function, managing complexity while maintaining precision.
Solution Approach 2:
The patent combines the wake-up word recognition and keyword recognition functions into a unified speech recognition system that shares common processing resources and algorithms. By merging these functions, the system achieves high recognition accuracy without proportionally increasing device complexity, as the same hardware components serve multiple recognition stages.
Data Source
AI summary
The present disclosure relates to a method and a device for recognizing speech in a vehicle. The method for recognizing the speech in the vehicle may include collecting one or more types of information, determining information to be linked with each other for speech recognition based on an information processing priority predefined corresponding to each type of the collected information, analyzing the determined information to perform the speech recognition for a signal input through a microphone, and extracting at least one of a wake up voice or a command voice through the speech recognition to control the vehicle. Therefore, the present disclosure has an advantage of more accurately performing the speech recognition by linking collected various information in the vehicle with each other.