Learned Filter Pair Response for Real-Time Signal Bandwidth Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing receiver systems face challenges in accurately and efficiently detecting and classifying repetitive signals in real-time due to high latency and inaccuracy in frequency and bandwidth estimation, especially in noisy environments.
Innovation Solution
The use of a learned filter pair with Infinite Impulse Response filters and machine learning techniques to estimate frequency and bandwidth, employing a grid of training data and 2-D interpolating lookup tables for real-time signal processing, enabling low-latency noise reduction and signal separation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If Fourier transform with backend signal detection and estimation processing is used, then frequency and bandwidth estimation can be performed, but the method has high latency and is not real-time
Solution Approach 1:
The patent divides the signal processing task into two distinct phases: an offline training phase where filter responses are pre-computed and stored in lookup tables, and an online estimation phase where pre-computed filters are applied to incoming signals. This segmentation allows computationally intensive operations to be performed beforehand, enabling real-time operation during actual signal processing while maintaining high estimation accuracy.
2Measurement precision
If signal energy estimates from three or more filters are used, then frequency and bandwidth estimation can be performed, but the method is not very accurate unless the signal is observed for long periods of time
Solution Approach 1:
The patent performs preliminary computation of optimal filter responses during an offline training phase, storing these pre-computed filters in lookup tables. During online operation, these pre-prepared filters are immediately applied to incoming signals, eliminating the need for long observation periods. The preliminary action of computing and storing optimal filters enables accurate real-time estimation without requiring extended signal monitoring.
3Reliability
If traditional filter methods are used for signal detection and separation, then signal processing can be performed, but the latency is too high for real-time operation
Solution Approach 1:
The patent creates pre-computed copies of optimal filter responses during an offline training phase and stores them in lookup tables. During online operation, these pre-computed filter copies are immediately applied to incoming signals without requiring real-time computation. This copying approach maintains reliable signal detection and separation capabilities while enabling low-latency real-time processing.
Data Source
AI summary
Systems and methods for estimating frequency and bandwidth of unknown signals using learned features of a filter pair for the purpose of detecting, separating and tracking these signals in an electronic receiver. The techniques could be part of a signal cueing system that initiates signal detection, separation and tracking or a signal separation and tracking system which is initialized by the cueing system and adaptively updates frequency and bandwidth estimates so as to continuously separate and track signals after initial detection. The methodology is to train the filter responses using machine learning by creating a grid of training data based on signal examples that cover a span of frequencies and bandwidths. The system estimates frequency and bandwidth in real time, inputs those estimates into interpolating lookup tables to retrieve filter coefficients, and provides those filter coefficients to a tunable tracking filter.