Microphone Speech Detection for Natural vs Replayed Voice
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing voice biometrics systems struggle to differentiate between natural and replayed speech, leading to high power consumption in battery-powered devices due to the inability of conventional voice activity detectors to distinguish between the two, which is particularly problematic in applications with limited power resources.
Innovation Solution
Implementing a low-power natural speech detection module in the analogue domain within a microphone to differentiate between natural and replayed speech by analyzing frequency bands and power ratios, reducing the need for power-intensive digital signal processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional voice activity detectors are used to detect speech, then speech processing can be activated, but power consumption increases significantly due to inability to distinguish between natural and replayed speech
Solution Approach 1:
The audio signal is divided into multiple frequency bands (first frequency band and second frequency band) for separate analysis. The natural speech detection module processes each band independently to extract specific characteristics, allowing the system to distinguish between natural and replayed speech more effectively while managing power consumption through targeted processing of only the necessary frequency components.
Solution Approach 2:
The system changes the analysis parameters by examining the power ratio between different frequency bands and detecting modulation characteristics at specific rates (4 Hz to 10 Hz). By transforming the detection approach to analyze these specific parameters, the system achieves better discrimination between natural and replayed speech without requiring full digital signal processing, thus reducing power consumption.
2Measurement precision
If power-intensive digital signal processing is used to detect replayed speech, then detection accuracy improves, but power consumption increases in battery-powered devices
Solution Approach 1:
The patent replaces complex digital signal processing mechanisms with an analogue-domain natural speech detection module. This substitution allows the system to perform speech detection using analogue circuitry that processes audio signals directly, avoiding the need for power-intensive digital processing while maintaining sufficient accuracy to distinguish between natural and replayed speech.
Solution Approach 2:
Instead of processing the entire audio signal through comprehensive digital analysis, the system applies partial action by focusing only on specific frequency bands (first and second frequency bands) and specific modulation rates (4 Hz to 10 Hz). This selective processing provides enough information for accurate detection without the excessive computational power consumption of full digital analysis.
3Productivity
If the natural speech detection module processes all audio signals, then detection coverage is maximized, but power consumption increases
Solution Approach 1:
The natural speech detection module operates dynamically by adjusting its processing based on the input signal characteristics. It activates and processes only when and where natural speech is detected, rather than continuously processing all audio signals. This dynamic operation maintains high detection coverage for actual speech events while minimizing power consumption during periods when no natural speech is present.
Data Source
AI summary
A microphone device, comprising: a microphone configured to generate an audio signal; a natural speech detection module for detecting natural speech in the first audio signal; wherein, on detection of the natural speech in the first audio signal, the natural speech detection module is configured to output a trigger signal to a speech processing module to process the natural speech in the audio signal.