Voice Operated Switch and Wake Word Detection Buffering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
When both Voice Operated Switch (VOX) and Voice Control with Wake Word features are active on a communications device, there is a challenge in distinguishing between intended voice commands and regular speech, leading to potential loss of communication as speech is buffered until the Wake Word is detected, causing delays and inefficiencies.
Innovation Solution
The device monitors for voice activity, buffers the speech, and upon detecting the Wake Word, transitions to Voice Control State, allowing subsequent speech to be processed as a voice command, with any speech not containing the Wake Word being transmitted over the communications channel after determining its absence, potentially using time compression to mitigate delay.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If speech is buffered until Wake Word is detected to distinguish voice commands from regular speech, then voice command recognition accuracy is improved, but communication delay increases
Solution Approach 1:
The system performs preliminary buffering of speech signals and pre-detection of Wake Words before full voice command processing is initiated. This allows the system to prepare in advance for potential commands while maintaining the ability to transmit regular speech promptly once the Wake Word is detected or buffering period expires.
Solution Approach 2:
The buffering duration and processing state are made dynamic rather than static. The system adjusts buffering behavior based on detected speech patterns, Wake Word presence, and current operational state, allowing flexible transition between command-mode buffering and regular speech transmission to optimize both accuracy and delay.
2Measurement precision
If speech is continuously buffered for Wake Word detection, then voice commands can be accurately identified, but speech transmission reliability deteriorates
Solution Approach 1:
The system implements a preliminary buffering phase with a defined timeout period before Wake Word detection is fully activated. This preliminary action ensures that regular speech without Wake Words can be transmitted after the buffering period expires, preventing permanent loss of communication while still allowing accurate Wake Word detection during the buffer period.
Solution Approach 2:
The system continuously monitors buffered speech for Wake Word patterns and provides feedback control on whether to continue buffering or switch to transmission mode. This feedback mechanism ensures that speech is not lost indefinitely and that the system transitions appropriately between command-detection state and regular communication state.
3Adaptability or versatility
If both VOX and Voice Control with Wake Word features are active simultaneously, then hands-free operation capability is improved, but system complexity increases
Solution Approach 1:
The system segments the voice processing functionality into distinct operational modes: VOX mode for immediate speech transmission and Voice Control with Wake Word mode for command processing. By segmenting these functions and managing them through state transitions rather than running them concurrently as independent complex systems, the overall system complexity is reduced while maintaining both capabilities.
Solution Approach 2:
The voice processing system is designed with multi-functionality to handle both VOX and Wake Word detection through a unified speech buffer and state machine architecture. This universal design allows the same hardware and software components to serve multiple purposes depending on the current operational state, reducing the need for separate dedicated systems for each feature.
Data Source
AI summary
A method is provided. The method includes enabling voice activated switch (VOX) mode on a communication device. The method further includes enabling a wake word detector on the communication device. The method also includes monitoring for voice activity using a voice activity detector (VAD). The method additionally includes detecting voice activity using the VAD, wherein detecting voice activity further includes: storing the voice activity in a buffer, determining, using the wake word detector, the presence or absence of a wake word in the voice activity, monitoring for and subsequently processing a voice command based on the determination of the presence of the wake word in the voice activity, and initiating a transmit operation, on the communication device, of the voice activity stored in the buffer based on the absence of the wake word in the voice activity.