Wearable Audio Source Identification via Phase Shift Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional wearable electronic devices require speech training and the use of specific trigger phrases for voice recognition, limiting their ability to respond to multiple users and making them impractical for devices like smart glasses used by multiple wearers without prior training.
Innovation Solution
The implementation of a speech analysis manager that uses multiple audio sensors to determine phase shifts in audio signals, allowing the device to identify speech generated by the wearer without the need for training or trigger phrases, by comparing measured phase shifts to expected values and eliminating non-wearer generated audio.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If speech training techniques are used to recognize user voice, then voice recognition accuracy is improved, but device complexity and ease of operation deteriorate due to requiring training procedures and trigger phrases
Solution Approach 1:
The system performs self-identification of the wearer by automatically analyzing audio phase shifts without requiring external training data or user-provided voice samples. The device serves itself by using its own audio sensors to detect and analyze the acoustic characteristics of speech originating from the wearer's position, eliminating the need for users to complete training procedures.
Solution Approach 2:
The patent replaces the conventional speech recognition system (which relies on trained voice models and trigger phrases) with a physics-based acoustic localization system. Instead of using machine learning models trained on user voice, the system uses phase shift analysis of audio waves captured by multiple sensors to geometrically determine the origin of speech, substituting a physical measurement approach for a data-driven recognition approach.
2Reliability
If speech training is required for voice recognition, then recognition reliability is improved, but adaptability deteriorates for multi-user devices like smart glasses
Solution Approach 1:
The phase shift-based speech origin detection system is universal and works for any wearer without requiring device-specific customization. The same hardware (multiple audio sensors) and algorithm (phase shift analysis) serve all users equally, making the device inherently adaptable to multiple users while maintaining consistent reliability across different wearers.
Solution Approach 2:
Instead of having the system adapt to each user through training (user-specific approach), the patent inverts the approach by having the system identify which user is speaking based on physical acoustic properties (phase shift) that are independent of user identity. This reversal allows the device to work with any user immediately without requiring adaptation or training procedures.
3Adaptability or versatility
If multiple audio sensors are used to determine phase shifts, then speech source identification capability is improved, but device complexity increases
Solution Approach 1:
The patent divides the audio detection function into multiple independent audio sensors positioned at different locations on the device. Each sensor independently captures audio signals, and the system processes each sensor's output separately through phase shift analysis. This segmentation enables spatial differentiation of speech sources while keeping each sensor's function simple and well-defined.
Solution Approach 2:
The patent introduces phase shift analysis as an intermediary processing step between audio capture and speech recognition. Instead of directly comparing raw audio signals from multiple sensors (which would be complex), the system uses phase shift as an intermediate physical quantity that simplifies the determination of speech origin. This intermediary approach transforms a complex multi-sensor problem into a manageable calculation based on wave propagation physics.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
Enables effortless voice detection for any user without the need for training or trigger phrases, making the technology suitable for multi-user applications and enhancing the natural interaction with wearable electronic devices.
Implementation Method 1
determine a phase shift between a first sample of first audio captured at a first audio sensor and a second sample of the first audio captured at the second audio sensor
Data Source
AI summary
Methods, systems and articles of manufacture for a wearable electronic device having an audio source identifier are disclosed. Example audio source identifiers disclosed herein include first and second audio sensors disposed at first and second locations, respectively, on a wearable electronic device. Such audio source identifiers also include a phase shift determiner to determine a phase shift between a first sample of first audio captured at the first audio sensor and a second sample of the first audio captured at the second audio sensor. The first audio includes first speech generated by a first speaker wearing the wearable electronic device. Example audio source identifiers further include a speaker identifier to determine, based on the phase shift determined by the phase shift determiner, whether second audio includes speech generated by a second speaker wearing the wearable electronic device.