Tonal Stability Estimation for Sound Activity Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current sound activity detection algorithms face challenges in accurately distinguishing between music and speech signals, leading to poor quality encoding and increased bit rates, especially in the presence of music signals, which affects the performance of existing codecs.
Innovation Solution
A method and device for estimating tonal stability in sound signals by calculating a current residual spectrum, detecting peaks, and calculating correlation maps between current and previous residual spectra to improve sound activity detection and background noise estimation, preventing false updates and enhancing music signal discrimination.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional sound activity detection algorithms are used, then speech signals can be encoded efficiently, but music signals are misclassified leading to poor encoding quality and increased bit rates
Solution Approach 1:
The patent introduces tonal stability as a new parameter to distinguish music from speech. By calculating the correlation between current and previous residual spectra, the system adds a temporal dimension to spectral analysis, enabling reliable differentiation between tonal music signals and speech while maintaining encoding efficiency
Solution Approach 2:
The residual spectrum acts as an intermediary representation that isolates the tonal components from the overall signal. By analyzing the correlation structure of this residual spectrum across time, the system creates a mediator mechanism that accurately identifies music signals without interfering with speech encoding performance
2Reliability
If background noise estimation is updated continuously, then noise adaptation improves, but false updates occur during music signals degrading performance
Solution Approach 1:
The system performs preliminary detection of music signals using tonal stability analysis before allowing background noise estimation updates. When music is detected, the system preemptively prevents false updates to the noise model, blocking the harmful effect before it can degrade performance
Solution Approach 2:
The tonal stability detection provides feedback to the noise estimation module, creating a control mechanism that adjusts update behavior based on signal type. This feedback loop enables the system to adapt noise estimation dynamically while preventing false updates during music segments
3Loss of energy
If VAD algorithms are used to reduce bit rate, then inactive frames are encoded efficiently, but music signals are misclassified as unvoiced speech resulting in severe quality degradation
Solution Approach 1:
The patent modifies the classification parameters by introducing tonal stability as a discriminative feature. This enables the VAD algorithm to maintain low bit rate encoding for truly inactive frames while preventing music signals from being misclassified, thus preserving encoding quality across different signal types
Data Source
AI summary
A device and method for estimating a tonal stability of a sound signal include: calculating a current residual spectrum of the sound signal; detecting peaks in the current residual spectrum; calculating a correlation map between the current residual spectrum and a previous residual spectrum for each detected peak; and calculating a long-term correlation map based on the calculated correlation map, the long-term correlation map being indicative of a tonal stability in the sound signal.