Vowel Sensing Voice Activity Detector for False Positive Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing voice activity detection systems based on sound level often generate false positives due to non-voice sounds, such as doors closing or keyboard typing, which are not accurately differentiated from human voice activity.
Innovation Solution
A vowel-based voice activity detection method that utilizes the harmonicity of human voice signals, specifically detecting signals with harmonic frequency components within a range of glottal pulse frequencies, and uses a circular autocorrelation of the absolute value of a short-time hamming windowed audio spectrum, normalized and filtered to reduce stationary noise, for accurate detection of human voice presence.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If sound level based VAD is used, then detection simplicity is maintained, but false positive rate increases due to non-voice sounds
Solution Approach 1:
The patent transforms the detection parameter from simple sound level to spectral harmonicity characteristics. By analyzing the harmonic structure of sounds in different frequency bands and comparing them against voice-like patterns, the system maintains operational simplicity while dramatically improving reliability and reducing false positives from non-voice sounds.
2Measurement precision
If vowel-based detection with harmonic analysis is used, then detection accuracy improves, but computational complexity increases
Solution Approach 1:
The patent divides the audio spectrum into multiple frequency bands and analyzes harmonicity independently in each band. This segmentation allows the complex harmonic analysis to be broken down into manageable parallel computations, improving detection accuracy while controlling computational complexity through distributed processing.
Solution Approach 2:
The patent focuses analysis only on specific frequency ranges where voice harmonics are most prominent (200Hz-4kHz), rather than analyzing the entire audio spectrum. This partial action approach concentrates computational resources on the most informative frequency regions, achieving high detection accuracy with reduced overall computational complexity.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Methods and apparatuses for detecting user speech are described. In one example, a method for detecting user speech includes receiving a microphone output signal corresponding to sound received at a microphone and identifying a spoken vowel sound in the microphone signal. The method further includes outputting an indication of user speech detection responsive to identifying the spoken vowel sound.