Bone-Conduction Voice Detection Using Time-Frequency Features
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing voice control systems in terminal devices face complexity and inefficiency in distinguishing voice signals from noise due to the need for simultaneous use of microphone and bone conduction sensor recognition processes.
Innovation Solution
A method utilizing a bone conduction sensor to detect voice signals by acquiring time and frequency domain features, such as short-term zero-crossing rate, pitch period, and spectral center of gravity, to determine voice presence without requiring microphone input.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If both microphone and bone conduction sensor are used for voice recognition, then voice detection accuracy is improved, but device complexity increases
Solution Approach 1:
The patent extracts and utilizes only the necessary features (zero-crossing rate and spectral centroid) from the bone conduction sensor signal for voice detection, eliminating the need to process signals from multiple sensors simultaneously. This reduces the recognition process complexity while maintaining voice detection accuracy by focusing on the most discriminative features.
Solution Approach 2:
The bone conduction sensor is made to serve multiple functions: it detects both voice signals and noise, and provides sufficient information for accurate voice detection without requiring additional dedicated sensors. This multi-functionality reduces device complexity by eliminating the need for separate microphone processing paths.
2Reliability
If both microphone and bone conduction sensor signals are processed simultaneously, then voice detection reliability is improved, but processing time increases
Solution Approach 1:
The patent extracts only the essential features (zero-crossing rate and spectral centroid) from the bone conduction sensor signal that are sufficient for reliable voice detection. By processing only these key features rather than full signal processing from multiple sensors, the recognition processing time is reduced while maintaining voice detection reliability.
Solution Approach 2:
The patent applies partial action by using only the necessary portion of the signal processing pipeline - specifically calculating zero-crossing rate and spectral centroid from bone conduction sensor data - rather than performing complete signal processing from multiple sensors. This partial processing approach reduces time loss while achieving sufficient detection reliability.
3Measurement precision
If multiple sensors and complex recognition processes are used, then voice detection accuracy is improved, but manufacturing cost increases
Solution Approach 1:
The bone conduction sensor is designed to perform multiple functions including voice detection, noise detection, and providing sufficient discriminatory information for accurate recognition. This multi-functionality reduces manufacturing cost by eliminating the need for additional dedicated voice detection sensors and their associated processing hardware.
Solution Approach 2:
The patent extracts and utilizes only the critical features from bone conduction sensor data that are sufficient for accurate voice detection. This approach reduces manufacturing cost by simplifying the processing architecture and eliminating the need for complex multi-sensor processing systems while maintaining detection accuracy.
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
Simplifies voice detection by relying solely on the bone conduction sensor, reducing costs and improving accuracy by distinguishing voice from noise more effectively.
Implementation Method 1
receiving a time domain signal detected by a bone conduction sensor in the terminal device
Data Source
AI summary
A voice signal detection method, a terminal device and a storage medium. Said method comprises: receiving a time domain signal detected by a bone conduction sensor in the terminal device, and acquiring time domain features in the time domain signal (S10); converting the time domain signal into a frequency domain signal, and acquiring frequency domain features in the frequency domain signal (S20); and when the time domain feature satisfies a first preset condition and the frequency domain feature satisfies a second preset condition, determining that a voice signal has been detected by the bone conduction sensor (S30). The voice detection is performed according to a signal detected by the bone conduction sensor, without the need of combining with a signal detected by a microphone, so that the voice detection is simpler, and moreover, as voice recognition is performed merely in combination with the bone conduction sensor, the cost is low.