Voice Input Apparatus Noise Isolation via Multi-Microphone Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing voice input apparatuses, such as wireless earphones with microphones, struggle to effectively transmit user voices to AI assistants in loud ambient noise environments, leading to poor voice recognition by AI assistants.
Innovation Solution
A voice input apparatus and method that utilize multiple microphones positioned at different locations relative to the speaker's mouth and ear canal, including a microphone outside the ear canal, one closer to the mouth, and one inside the ear canal, to collect and process voice signals. The apparatus includes a controller that detects sound pressure, sets reflection degrees for the collected voice signals, and generates an output voice signal for transmission to an AI assistant.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single microphone is used to collect voice in loud ambient noise, then the device structure is simple, but the voice recognition accuracy deteriorates
Solution Approach 1:
The voice collection function is segmented into three separate microphones positioned at different locations: one inside the ear canal, one at the first position outside the ear canal, and one at the second position closer to the mouth. Each microphone captures different acoustic characteristics, and the controller selectively uses or combines these signals based on ambient noise conditions to improve voice recognition accuracy.
2Measurement precision
If multiple microphones are used to improve voice recognition in noise, then the voice recognition accuracy is improved, but the device complexity increases
Solution Approach 1:
Each microphone is assigned a specific local position with distinct acoustic properties: the first microphone inside the ear canal captures bone conduction sounds and ambient noise, the second microphone at the first position outside the ear canal captures ambient noise, and the third microphone at the second position closer to the mouth captures direct voice signals. The controller applies local quality processing by selectively choosing or combining signals from specific microphones based on the acoustic environment, optimizing voice recognition while managing device complexity.
3Loss of information
If all microphone signals are transmitted to AI assistant, then the voice signal is complete, but the noise interference increases
Solution Approach 1:
The controller extracts and removes noise components from the microphone signals before transmitting to the AI assistant. It compares signals from multiple microphones, identifies ambient noise patterns (especially from the first microphone inside the ear canal and second microphone outside), and selectively extracts only the useful voice components. This extraction process eliminates harmful noise factors while preserving complete voice information for accurate AI recognition.
4Measurement precision
If noise filtering is applied to improve signal quality, then the voice recognition accuracy is improved, but the processing complexity increases
Solution Approach 1:
The signal processing approach is made dynamic rather than static. The controller adaptively adjusts which microphone signals to use or combine based on real-time acoustic environment assessment. In high ambient noise conditions, it dynamically switches to prioritize signals from microphones less affected by noise or applies dynamic noise cancellation algorithms. This dynamic processing optimizes voice recognition accuracy while avoiding unnecessary processing complexity in quiet environments.
Data Source
AI summary
A voice input apparatus includes first to third microphones and a controller. The first microphone collects a voice at a first position outside an ear canal of a speaker, and outputs a first input voice signal. The second microphone collects a voice at a second position outside the ear canal of the speaker and closer to a mouth of the speaker than the first position, and outputs a second input voice signal. The third microphone collects a voice inside the ear canal of the speaker, and outputs a third input voice signal. The controller detects a sound pressure of the first input voice signal, sets reflection degrees of the second and third input voice signals according to the detected sound pressure, and generates an output voice signal including at least one of the second and third input voice signals based on the reflection degrees.