Earphone Voice Recognition Audio Buffering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Voice recognition systems, such as those in earphones, often experience word loss and unreliable recognition due to the time delay between detecting mouth movement and picking up sound, leading to incomplete wake-word detection.
Innovation Solution
A method for voice recognition in earphones that buffers and combines audio data received before and after a mouth movement is detected, ensuring complete wake-word capture by defining specific durations for data reception and processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the voice recognition system waits to detect mouth movement before recording audio data, then the recognition accuracy is improved, but words will be lost during the detection time delay
Solution Approach 1:
The system performs preliminary action by starting to buffer audio data before the mouth movement detection is complete. When the detection triggers, the audio data has already been captured during the detection period, preventing any word loss while maintaining recognition accuracy.
2Loss of information
If the system starts recording immediately without waiting for mouth movement detection, then word loss is prevented, but recognition reliability decreases due to incomplete wake-word detection
Solution Approach 1:
The system performs preliminary action by starting to buffer audio data before the mouth movement detection is complete. When the detection triggers, the audio data has already been captured during the detection period, preventing any word loss while maintaining recognition accuracy.
3Loss of information
If the system buffers audio data during the detection period, then word loss is prevented, but the device complexity increases
Solution Approach 1:
The system merges the audio buffering function with the existing detection mechanism. The audio data buffered during the detection period is seamlessly integrated with the recognition process, preventing word loss without requiring a separate complex subsystem.
Data Source
Figure 1~2
Figure 3~4
Figure 5~6
AI summary
A method for voice recognition via an earphone is disclosed. The method includes receiving first audio data via the first microphone and buffering the first audio data in response to the first trigger signal; receiving second audio data via the first microphone and recognizing whether the first audio data contains data of a wake-on-voice word in response to the second trigger signal; and recognizing whether the second audio data contains data of the wake-on-voice word. The first audio data is received and buffered in a first duration starting from when the first trigger signal is received and ending when the second trigger signal is received. The second audio data is received in a second duration starting from when the second trigger signal is received and ending when whether the first audio data contains data of the wake-on-voice word is recognized.