Audio Classification Based on Perceptual Quality
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing audio classification methods at low or medium bit rates fail to accurately distinguish between audio and voiced signals, leading to suboptimal encoding in either the time-domain or frequency-domain, which affects perceptual quality.
Innovation Solution
The method re-classifies audio signals as voiced when periodicity parameters exceed certain thresholds, allowing for time-domain encoding, while maintaining frequency-domain encoding for audio signals, thereby improving perceptual quality by leveraging Long-Term Prediction gains in CELP coding.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of substance
If audio signals are classified as AUDIO and encoded in frequency-domain, then data compression is achieved, but perceptual quality deteriorates for periodic signals
Solution Approach 1:
The patent applies dynamic classification by analyzing periodicity parameters (pitch correlation, pitch difference) to determine whether each audio signal should be encoded in time-domain or frequency-domain. This dynamic adaptation allows the system to switch encoding modes based on signal characteristics, resolving the contradiction between compression efficiency and perceptual quality for periodic signals.
Solution Approach 2:
The patent changes the classification parameter from simple audio detection to multi-parameter periodicity analysis including pitch correlation values and pitch differences. By monitoring these parameters and comparing them against thresholds, the system identifies periodic signals that would benefit from time-domain encoding, thus improving perceptual quality while maintaining compression for non-periodic signals.
2Manufacturing precision
If audio signals are re-classified as VOICED and encoded in time-domain, then perceptual quality improves for periodic signals, but classification accuracy worsens without proper periodicity analysis
Solution Approach 1:
The patent performs preliminary analysis of periodicity parameters (normalized pitch correlation values and pitch differences between subframes) before making classification decisions. This preliminary action ensures that re-classification from AUDIO to VOICED is based on accurate measurement of signal periodicity, thereby maintaining classification accuracy while enabling appropriate time-domain encoding for periodic signals.
Solution Approach 2:
The patent replaces simple audio/voiced classification mechanisms with a more sophisticated system that substitutes mechanical classification with perceptual quality-based classification using periodicity parameters. This substitution allows the system to accurately identify periodic signals that require time-domain encoding, improving both classification accuracy and perceptual quality.
3Speed
If simple audio classification is used, then processing speed is maintained, but encoding optimality worsens
Solution Approach 1:
The patent applies partial action by selectively analyzing periodicity parameters only for signals that require further classification, rather than applying complex analysis to all audio signals. This approach maintains processing speed for straightforward cases while achieving encoding optimality for periodic signals that require time-domain encoding, thus resolving the contradiction between speed and optimality.
Data Source
AI summary
The quality of encoded signals can be improved by reclassifying AUDIO signals carrying non-speech data as VOICE signals when periodicity parameters of the signal satisfy one or more criteria. In some embodiments, only low or medium bit rate signals are considered for re-classification. The periodicity parameters can include any characteristic or set of characteristics indicative of periodicity. For example, the periodicity parameter may include pitch differences between subframes in the audio signal, a normalized pitch correlation for one or more subframes, an average normalized pitch correlation for the audio signal, or combinations thereof. Audio signals which are re-classified as VOICED signals may be encoded in the time-domain, while audio signals that remain classified as AUDIO signals may be encoded in the frequency-domain.