Multi-Microphone Wake-Word Detection for In-Car Voice Isolation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional voice recognition systems in vehicles fail to accurately detect the direction of a wake-up word in noisy environments, leading to delayed activation of vehicular devices due to the inability to differentiate between voices from different directions.
Innovation Solution
A voice processing device with multiple microphones arranged across seats, a storing unit, a word detection unit, a microphone determining unit, and a voice processing unit that stores voice signals, detects the presence of a wake-up word, determines the speaker's position, and suppresses other voices to isolate the speaker's voice.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple microphones are used to detect voice direction in noisy vehicle environments, then voice recognition accuracy is improved, but device complexity increases
Solution Approach 1:
The vehicle interior is divided into multiple detection zones, each assigned to a specific microphone. The microphones are strategically positioned at different locations (front, rear, left, right) to segment the monitoring space, allowing independent analysis of voice signals from each direction while maintaining overall system accuracy.
Solution Approach 2:
Multiple microphone signals are merged and processed together through signal processing algorithms. The system combines the output from all microphones to create a comprehensive voice detection result, leveraging the collective information from all sensors to improve recognition accuracy while managing complexity through integrated processing.
2Reliability
If voice signals from all directions are processed equally, then no voice is lost, but activation speed decreases due to inability to prioritize relevant voices
Solution Approach 1:
Different processing priorities are assigned to different microphone signals based on their spatial location and relevance. The system applies local quality control by giving higher weight to microphones detecting voices from the driver's direction or front seats, while still monitoring all other directions to ensure no important voice is missed.
Solution Approach 2:
The system continuously monitors voice signals from all microphones and provides feedback to adjust processing priorities in real-time. When a wake-up word is detected from a specific direction, the system feedback-adjusts the processing focus to that direction while maintaining awareness of other directions, enabling fast activation without losing comprehensive monitoring.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
Enables quick detection and extraction of the wake-up word's direction, enhancing the accuracy and speed of voice recognition, and allowing for timely activation of vehicular devices.
Implementation Method 1
a plurality of different microphones MC1-MC6 corresponding to a plurality of respective seats
Implementation Method 2
a voice processing unit which outputs a voice uttered by the speaker while suppressing a voice uttered by a passenger other than the speaker
Data Source
AI summary
A voice processing device includes plural microphones arranged so as to correspond to a plurality of positions. The voice processing device includes at least one memory that stores instructions and voice signals from the plural microphones, and a processor. The voice signals collected by the plural microphones, respectively, during a prescribed period before a present time, are repeatedly stored in the at least one memory as buffered voice signals. The processor detects whether a prescribed word is uttered by a speaker based on the voice signals collected by the plural microphones, determines a microphone corresponding to the speaker by referring to the buffered voice signals, and suppresses the voice signals collected by the plural microphones other than the microphone corresponding to the speaker.