Signal Analysis Filter Bank for Instantaneous Frequency Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current time-frequency analysis methods, such as the spectrogram associated with the windowed Fourier Transform, face limitations due to the uncertainty principle, which restricts simultaneous accuracy in time and frequency, leading to unclear and inaccurate results, especially for low-frequency components in audio signal processing.
Innovation Solution
A method using a filter bank with a plurality of frequency selective filters operating on distinct center frequencies, performing frequency translation and selective filtering to extract instantaneous phase, amplitude, and frequency information, allowing for continuous and accurate time-frequency analysis comparable to reassignment methods, suitable for low-cost computing platforms.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If windowed Fourier transform with time domain windows is used, then frequency domain information can be extracted, but the transitions between stationary periods cannot be seen well due to the stationarity assumption
Solution Approach 1:
The patent divides the frequency spectrum into multiple frequency bands and processes each band separately through parallel filter banks. This segmentation allows the system to track frequency components continuously without requiring stationarity assumptions, thereby capturing transitions between stationary periods while maintaining frequency detection accuracy.
Solution Approach 2:
The patent employs dynamic filter banks that continuously adapt to signal changes rather than assuming stationarity. The filters operate in parallel and can track frequency components in real-time, enabling the system to detect transitions between stationary periods while maintaining accurate frequency measurement.
2Measurement precision
If reassignment method is used for time-frequency analysis, then clarity and accuracy of frequency extraction is improved, but computing cost increases
Solution Approach 1:
The patent replaces the complex mathematical operations of the reassignment method with a parallel filter bank architecture. Instead of performing intensive post-processing calculations to achieve sharp frequency localization, the system uses multiple parallel filters tuned to different frequencies, each providing direct frequency measurement. This substitution maintains high frequency extraction accuracy while reducing computing requirements to levels suitable for low-end microcontrollers and FPGAs.
Solution Approach 2:
The patent segments the frequency analysis into parallel filter banks, where each filter handles a specific frequency band. This segmentation distributes the computational load across multiple simple parallel operations rather than requiring complex sequential processing, thereby achieving reassignment-like accuracy on resource-constrained devices.
3Productivity
If audio filter banks operating in parallel are used, then frequency domain information can be extracted efficiently, but the accuracy for low frequencies below 800 Hz is insufficient
Solution Approach 1:
The patent applies local quality by using different filter configurations for different frequency ranges. For low frequencies below 800 Hz, the system employs filters with narrower bandwidths and optimized characteristics to achieve higher precision, while higher frequencies use broader bandwidths for efficient processing. This localized optimization ensures high accuracy for low frequencies without sacrificing overall processing efficiency.
Solution Approach 2:
The patent implements dynamic filter banks that can adjust their characteristics based on the input signal properties. For low-frequency components, the filters dynamically adapt to provide enhanced resolution and accuracy, while maintaining efficient parallel processing across the entire frequency spectrum. This dynamic adaptation allows the system to achieve high precision for low frequencies without compromising productivity.
Data Source
AI summary
The present invention is relative to a method for analyzing an signal (INS), representative of a wave that propagates in a physical medium, providing characteristic parameters of said signal, said method being implemented on a computing platform (CP), requiring only fixed point computations, and with a reduced number of multiplications. Parameters that are provided can be one or several of the following: instantaneous phase (IP), instantaneous amplitude (IA), instantaneous frequency (IF), Sliding Fourier Transform (STFT).