Speech-Pose Fusion Wake Detection to Prevent False Voice Activation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The challenge of reducing the probability of speech assistants being accidentally activated without a wake-up word in speech interactions, leading to user inconvenience and reduced experience.
Innovation Solution
A method involving a speech detection model, pose detection model, and a speech-pose detection fusion model to determine the likelihood of a valid speech instruction, using confidence levels to accurately activate the speech assistant only when intended by the user.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If speech detection is performed without a wake-up word, then speech interaction becomes more natural and user-friendly, but the probability of accidental activation increases
Solution Approach 1:
The patent introduces a new dimension of detection by combining speech detection with pose detection. Instead of relying solely on speech signal analysis, the system adds spatial information from accelerometer sensors to determine device orientation. This multi-dimensional approach allows the system to distinguish between intentional speech commands and accidental activations by checking both the speech content and the device's physical state.
Solution Approach 2:
The system dynamically adjusts the activation threshold by incorporating multiple parameters: speech confidence level, pose confidence level, and device state information. Rather than using a single fixed threshold for speech recognition, the system evaluates multiple parameters simultaneously and adjusts the overall activation decision based on the combination of these parameters, making the activation criteria more precise and adaptable.
2Reliability
If multiple detection models are used to reduce accidental activation, then activation accuracy improves, but system complexity increases
Solution Approach 1:
The patent divides the detection system into separate modular components: a speech detection model for analyzing audio signals, a pose detection model for analyzing accelerometer data, and a fusion model for combining their outputs. Each model operates independently and processes specific types of data, making the overall system more manageable and maintainable despite its complexity. This segmentation allows for independent optimization and debugging of each component.
Solution Approach 2:
The system merges the outputs of multiple detection models through a fusion model that combines speech confidence levels with pose confidence levels. This merging approach allows the system to leverage the strengths of each individual model while achieving more accurate overall detection. The fusion model integrates information from both speech and pose detection to make the final activation decision, reducing false positives without requiring a completely new complex system.
Data Source
AI summary
This application provides a speech interaction method and a related electronic device. The method includes: receiving a first speech signal; obtaining speech signal data based on the first speech signal when it is determined that speech detection is to be performed on the first speech signal; processing the speech signal data by using a speech detection model to obtain a first confidence level; obtaining pose information of the electronic device based on the acceleration data; processing the pose information by using a pose detection model to obtain a second confidence level; processing target pose information and speech data by using an speech pose detection fusion model to obtain a third confidence level; and determining, based on the three confidence levels, whether to start a speech interaction application. According to the method, the speech interaction application of the electronic device may be prevented from being woken up by mistake.