Hardware Abstraction Layer Speech Wake-Up Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current speech wake-up technologies experience significant delays due to long audio data transmission links, leading to late wake-up times and impaired decision speed in recognizing wake-up phrases.
Innovation Solution
The method involves obtaining audio data through an audio-video interface on a hardware abstraction layer, performing noise reduction processing, and conducting wake-up phrase detection within the same software layer, eliminating the need for transmission to the application layer, thereby reducing wake-up delay and improving efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If audio data is transmitted through the application layer for wake-up detection, then the system follows standard software architecture, but wake-up delay increases and decision speed decreases
Solution Approach 1:
The patent segments the wake-up detection function from the application layer and places it in the hardware abstraction layer, separating the time-critical detection task from the general-purpose application software. This segmentation allows the detection to occur at a lower level where it can access audio data directly without going through the application layer transmission link, thereby reducing wake-up delay while maintaining system architecture integrity.
Solution Approach 2:
The patent changes the dimensional layer of wake-up detection from the application layer to the hardware abstraction layer, moving the detection function to a different software layer that is closer to the hardware. This dimensional change enables direct access to audio data at the hardware abstraction layer, eliminating the need for transmission through the application layer and reducing the length of the audio data transmission link.
2Measurement precision
If noise reduction processing is performed before wake-up detection, then speech recognition accuracy improves, but processing time increases
Solution Approach 1:
The patent performs noise reduction processing as a preliminary action before wake-up detection by placing both operations in the hardware abstraction layer. This preliminary processing of audio data removes background noise and enhances speech signals before they are analyzed for wake-up phrases, improving speech recognition accuracy. Since both operations occur in the same layer, the processing is optimized and efficient, minimizing the time penalty.
Solution Approach 2:
The patent merges the noise reduction processing and wake-up detection operations into the same software layer (hardware abstraction layer). This merging allows both operations to share the same processing pipeline and resources, enabling noise reduction to be performed concurrently with or immediately before wake-up detection without requiring separate processing stages. The combined approach improves speech recognition accuracy while optimizing processing time through efficient resource utilization.
Data Source
AI summary
The present disclosure provides a method and an apparatus for waking up via a speech. The method includes: obtaining audio data currently collected by an electronic device through an audio-video interface on a hardware abstraction layer of the electronic device; performing noise reduction processing on the collected audio data to generate audio data to be detected; performing wake-up phrase detection on the audio data to be detected to determine whether the audio data to be detected includes a wake-up phrase; and performing wake-up processing on the electronic device in response to determining that the audio data to be detected comprises the wake-up phrase.