Personalized Voice Activity Detection Using Statistical Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing voice activity detection systems face challenges in accurately and efficiently isolating the voice of a specific speaker in real-time, leading to high error rates and unacceptable conversation quality due to delayed processing.
Innovation Solution
A system that registers user voice patterns and analyzes audio data segment by segment in real-time, calculating a probability level to determine if the segment contains the voice of a registered user, and selectively transfers or adjusts the signal based on this determination, optionally blocking or transferring segments below a threshold.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If statistical methods are used to determine voice presence and speaker identity in real-time, then processing can be performed continuously, but the determination accuracy is insufficient and requires further evaluation that introduces delay
Solution Approach 1:
The audio signal is divided into segments, and each segment is evaluated independently using statistical methods. This allows continuous processing while maintaining the ability to make precise determinations about voice presence and speaker identity in each segment, resolving the contradiction between continuous processing and determination accuracy.
2Measurement precision
If delay is introduced to improve determination accuracy, then more accurate speaker identification is achieved, but conversation quality becomes unacceptable
Solution Approach 1:
The system performs statistical analysis on each audio segment independently, making determination decisions based on partial evaluation of each segment rather than waiting for complete analysis of the entire signal. This partial action approach maintains high accuracy for each segment while avoiding excessive delay that would degrade conversation quality.
3Speed
If instantaneous transfer of audio is performed, then real-time communication is maintained, but high error rates occur in voice isolation
Solution Approach 1:
The system performs statistical analysis and voice isolation determination on each audio segment before transferring it, using preliminary evaluation to identify and isolate the speaker's voice. This preliminary action enables real-time transfer while maintaining high reliability in voice isolation through pre-computed statistical determinations.
4Measurement precision
If sophisticated input devices are used to differentiate speakers and noise, then voice isolation improves, but device complexity increases
Solution Approach 1:
The system replaces physical sophisticated input devices with statistical analysis methods that process audio segments using computational algorithms. This substitution achieves high speaker differentiation capability through software-based statistical evaluation rather than through additional hardware, thereby improving measurement precision while reducing device complexity.
Data Source
AI summary
A method of transferring a real-time audio signal transmission, including: registering voice patterns (or other characteristics) of on more users to be used to identify the voices of the users, accepting an audio signal as it is created as a sequence of segments, analyzing each segment of the accepted audio signal to determine if it contains voice activity (314), determining a probability level that the voice activity of the segment is of a registered user (320&322); and selectively transferring the contents, of a segment responsive to the determined probability level (324).