Location-Aware Power Management for Always-On Voice Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Always-on-always-listen voice recognition systems in mobile devices face significant power consumption issues due to false triggers, which reduce battery life and user experience, as they often enter an active 'Key Phrase Detection' state unnecessarily, consuming higher power.
Innovation Solution
Implementing a location-aware power management scheme that uses sensors to automate the setting of ambient noise trigger profiles, reducing the frequency of active 'Key Phrase Detection' state and minimizing false triggers by calibrating noise thresholds based on location, time, and ambient conditions, thereby optimizing power consumption and user experience.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If the voice recognition system operates in always-on-always-listen mode with continuous Key Phrase Detection, then the system can detect voice commands immediately, but power consumption increases significantly
Solution Approach 1:
The patent segments the voice detection process into two distinct stages: a low-power first stage that continuously monitors for potential voice commands using simple threshold comparison, and a second stage that activates only when the first stage detects a potential match. This segmentation allows the system to maintain fast detection capability while minimizing the time the high-power processing components remain active.
Solution Approach 2:
The system employs periodic action by having the first stage continuously monitor audio input at low power consumption, and only activating the second stage periodically when a potential voice command is detected. This periodic activation of the high-power processing stage significantly reduces overall power consumption while maintaining responsive voice command detection.
2Measurement precision
If the system uses a low noise trigger threshold to detect voice commands in quiet environments, then detection sensitivity improves, but false triggers increase in noisy environments
Solution Approach 1:
The patent implements dynamic noise threshold adjustment where the trigger threshold is not fixed but adapts based on the detected ambient noise level. The first stage continuously monitors the audio environment and dynamically adjusts the trigger threshold to maintain optimal detection sensitivity across varying noise conditions, preventing both missed detections and false triggers.
Solution Approach 2:
The system changes the noise threshold parameter dynamically based on ambient conditions. By monitoring the audio environment and adjusting the trigger threshold parameter accordingly, the system maintains high detection sensitivity in quiet environments while preventing false triggers in noisy environments, thus resolving the contradiction between sensitivity and reliability.
3Reliability
If the system enters active Key Phrase Detection state frequently to ensure no voice commands are missed, then detection reliability improves, but battery life decreases
Solution Approach 1:
The patent segments the detection system into a continuously active low-power first stage and an intermittently active second stage. The first stage ensures no voice commands are missed by continuously monitoring audio input, while the second stage remains in lower-power states unless activated by the first stage, thus maintaining detection reliability while extending battery life.
Solution Approach 2:
The first stage acts as an intermediary between the continuous audio input and the high-power second stage processing. It filters and pre-processes the audio signal, activating the second stage only when necessary, thereby ensuring reliable detection while minimizing the time the power-consuming components remain active.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Methods and apparatus relating to a location aware power management scheme for an always-on-always-listen voice recognition system are described.In an embodiment,logic performs ambient noise trigger level analysis for a location and causes storage of an ambient noise trigger level threshold value for the location based on the ambient noise trigger level analysis. Furthermore,logic determines whether to cause modification to a state of an audio processor in response to detection of the audio processor at the location and comparison of a detected sound level at the location and the stored ambient noise trigger level threshold value. Other embodiments are also disclosed and claimed.