Voice Playback Speed Conversion via Autocorrelation Zero Cross Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for converting voice data playback speed are computationally intensive, requiring high-performance CPUs, making real-time processing challenging, especially in portable devices, and are prone to quality deterioration due to varying basic frequencies in human voices.
Innovation Solution
A method involving DC component removal, low-pass filtering, extraction of rising zero cross points, and autocorrelation calculations to segment voice blocks, allowing for efficient playback speed conversion even in low-performance CPUs without deteriorating quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If high-performance CPU is used to perform voice data playback speed conversion, then processing quality is improved, but device portability and battery capacity are reduced
Solution Approach 1:
The patent changes the computational parameters by using autocorrelation functions to determine voice block boundaries based on periodicity detection rather than complex spectral analysis. This parameter change enables accurate voice block segmentation using low-performance CPUs, resolving the contradiction between processing quality and device portability.
Solution Approach 2:
The patent replaces complex mechanical signal processing systems with mathematical autocorrelation calculations. By substituting complex spectral analysis and signal processing algorithms with autocorrelation-based periodicity detection, the system achieves comparable or superior performance with minimal computational resources, enabling portable device operation.
2Manufacturing precision
If complex calculation is performed to accurately extract voice blocks, then playback quality is improved, but processing time increases
Solution Approach 1:
The patent extracts only the essential periodicity information from voice signals using autocorrelation functions, discarding unnecessary spectral and temporal details. By taking out only the critical periodic characteristics needed for voice block boundary detection, the system achieves high extraction accuracy with minimal processing time.
Solution Approach 2:
The patent segments the voice signal processing into discrete autocorrelation calculation steps at regular intervals rather than continuous complex analysis. This segmentation approach maintains accurate voice block extraction by evaluating periodicity at key moments, significantly reducing processing time while preserving quality.
3Device complexity
If uniform processing is applied to all voice signals, then processing simplicity is improved, but accuracy deteriorates due to varying basic frequencies
Solution Approach 1:
The patent introduces dynamic adaptation by calculating autocorrelation functions that automatically adjust to the specific periodicity of each voice signal. Rather than using fixed uniform parameters, the system dynamically determines voice block boundaries based on the actual periodic characteristics detected through autocorrelation, achieving high accuracy across varying basic frequencies while maintaining processing simplicity.
Data Source
AI summary
The present invention addresses the problems of enabling a process of converting voice data playback speed even in a voice data playback device alone. The solution is a voice data playback speed conversion method and a voice data playback speed conversion device, comprising: a step of setting a reference zero cross point from any arbitrary zero cross point; a step of selecting a zero cross point temporally after the reference zero cross point within a first predetermined time range; a step of calculating a reference correlation function in a waveform from the reference zero cross point until a second predetermined time; and a step of calculating a correlation function in a waveform from a plurality of previously selected zero cross points until the second predetermined time, wherein a second reference zero cross point is the zero cross point of the waveform having a correlation function in which a concordance rate of the correlation value between the reference correlation function and the correlation function is the highest value, the difference between the reference zero cross point and the second reference zero cross point is calculated as a basic cycle, and the expansion and contraction of voice data is executed in basic cycle units so as to perform a process of converting the playback speed of the voice data.