Complex Data Encoding with Frequency-Domain Gap Compression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for encoding and compressing complex data, such as IQ samples for wireless communication, are inefficient for signals with frequency-domain gaps and require high computational complexity due to aggressive resampling near the Nyquist limit.
Innovation Solution
A system and method involving a processor-based encoder and decoder that apply window functions, discrete Fourier-related transforms, quantization, and overlapping time frames to encode and decode complex data, reducing computational complexity and improving compression efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If aggressive resampling close to the Nyquist limit is applied to compress fronthaul signals, then compression efficiency is improved, but computational complexity increases due to long FIR filters
Solution Approach 1:
The patent transforms the signal from time domain to frequency domain using Fourier-related transforms, changing the representation parameters of the signal. This allows compression to be applied in the frequency domain where spectral gaps are explicitly visible, avoiding the need for aggressive time-domain resampling and long FIR filters while achieving similar or better compression efficiency.
Solution Approach 2:
The patent extracts and removes frequency components corresponding to spectral gaps in the frequency domain representation. By identifying and eliminating unused frequency bands where no signal energy exists, the system achieves compression without requiring complex time-domain filtering operations.
2Productivity
If resampling to a lower sampling frequency is applied, then data compression is achieved, but effectiveness is reduced for signals with frequency-domain gaps
Solution Approach 1:
The patent changes the domain of signal representation from time domain to frequency domain through Fourier-related transforms. This parameter change enables direct visualization and exploitation of frequency-domain gaps, allowing the compression algorithm to target only the actual signal-bearing frequency components while skipping empty spectral regions, thereby improving effectiveness for signals with frequency gaps.
Solution Approach 2:
The patent transitions the signal analysis from one-dimensional time-domain sampling to two-dimensional frequency-domain representation. This dimensional change reveals the spectral structure of the signal, making frequency gaps explicitly visible and accessible for targeted compression, thus improving effectiveness compared to conventional time-domain resampling approaches.
3Productivity
If overlapping time frames are used in encoding, then compression efficiency is improved, but decoding complexity increases due to overlap-add operations
Solution Approach 1:
The patent applies periodic windowing functions to segmented time frames of the signal, creating a structured periodic processing pattern. The overlapping windows are designed with specific properties (such as raised cosine windows) that enable efficient reconstruction through simple addition operations, balancing compression efficiency with manageable decoding complexity.
Data Source
AI summary
It is provided a method for encoding complex data. The method includes the steps of: obtaining an input signal made up of a series of numerically represented samples; determining a time frame of the input signal to process; applying a first window function on data in the time frame, resulting in first windowed data, wherein the first window function tapers sample magnitude towards the edges of the first window function; performing a windowed complex discrete Fourier-related transform on the first windowed data, resulting in frequency-domain data including a plurality of coefficients, keeping only the real part or the imaginary part of each coefficient; quantizing the frequency domain data resulting in quantized data; outputting the quantized data as encoded data; and repeating the method, wherein each subsequent iteration of the step of determining a time frame includes determining a time frame that overlaps in time with a preceding time frame.