Logarithmic Frequency Spectrogram for Robust Fundamental Frequency Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for calculating fundamental frequency change in speech signals are prone to errors due to background noise, especially when the range of the fundamental frequency is not suitably set, leading to incorrect acquisition of the fundamental frequency change.
Innovation Solution
An apparatus and method utilizing a spectrogram calculation unit to generate a logarithmic frequency spectrum, a Hough transform unit to detect straight lines representing harmonic structures, and a change calculation unit to extract the fundamental frequency change without explicitly extracting the fundamental frequency, thereby reducing the influence of background noise.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the autocorrelation function of the predicted residual is used to calculate the fundamental frequency change, then the extraction error of the fundamental frequency is reduced, but the calculation becomes sensitive to background noise
Solution Approach 1:
The patent transforms the one-dimensional autocorrelation function into a two-dimensional logarithmic frequency spectrogram by applying logarithmic frequency transformation. This dimensional change allows the use of Hough transform to detect straight lines representing harmonic structures, making the fundamental frequency change calculation more robust against background noise while maintaining extraction accuracy.
Solution Approach 2:
The patent introduces the logarithmic frequency spectrogram as an intermediary representation between the predicted residual and the fundamental frequency change calculation. This intermediary transforms the signal into a domain where harmonic structures appear as straight lines, enabling more reliable detection through Hough transform even in noisy environments.
2Adaptability or versatility
If the range of the autocorrelation function is not limited to a specific fundamental frequency range, then the method can be more universally applicable, but the calculation becomes more susceptible to background noise
Solution Approach 1:
By transforming to logarithmic frequency domain and creating a spectrogram, the patent enables detection of harmonic structures across any fundamental frequency range without pre-specifying the range. The Hough transform naturally identifies straight lines regardless of their position in the frequency domain, providing universal applicability while maintaining noise robustness.
Solution Approach 2:
The logarithmic frequency spectrogram approach with Hough transform serves multiple functions: it detects harmonic structures across any fundamental frequency range, identifies the fundamental frequency change through straight line gradients, and remains robust against background noise. This multi-functional approach eliminates the need for range limitation while maintaining noise resistance.
Data Source
AI summary
A logarithmic frequency spectrum within a predetermined time range is calculated from a speech signal. The logarithmic frequency spectrum has a frequency element at equal intervals along a logarithmic frequency axis. A logarithmic frequency spectrogram is calculated by connecting a plurality of logarithmic frequency spectrums. A value of the frequency element along a straight line on the logarithmic frequency spectrogram is voted onto a Hough plane. The Hough plane has a voted value in correspondence with a gradient of the straight line. The voted value above a threshold and the gradient corresponding to the voted value are extracted from the Hough plane. A fundamental frequency change is calculated using the voted value and the gradient extracted.