Radar Sequence Reconstruction Using FFT for Dropped Frames

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Radar systems in mobile devices face challenges with limited datasets, random frame dropping, and mixed noise and target motion portions in range-Doppler images, which hinder effective data augmentation and sequence restoration.

Innovation Solution

A computer-implemented method using deep learning techniques, including FFT and IFFT operations, to restore sequence order, identify and remove noise, and augment data by decoding time domain information, utilizing auto-encoders and convolutional neural networks to process radar data.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If frame dropping occurs in radar data collection, then data collection efficiency is improved, but sequence completeness deteriorates

Engineering Contradiction:

Improvedata collection efficiencyVSAvoidsequence completeness

Core Design Contradiction:

ProductivityVSLoss of information

Solution Approach 1:

The patent applies preliminary action by performing FFT transformation and noise filtering on the frequency domain data before the actual sequence restoration is needed. The auto-encoder model is pre-trained to recognize and reconstruct missing frames, so when frame dropping occurs, the system can immediately restore the complete sequence without delay, thus maintaining both efficiency and completeness.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent introduces frequency domain representation as an intermediary between the dropped frames and the restored sequence. By transforming time-domain data to frequency domain using FFT, processing in the frequency domain, and then transforming back using IFFT, the system can recover missing information that would be lost in direct time-domain processing, thus resolving the contradiction between efficiency and completeness.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Quantity of substance

If noise portions are retained in the dataset, then data volume is maintained, but classification accuracy deteriorates

Engineering Contradiction:

Improvedata volumeVSAvoidclassification accuracy

Core Design Contradiction:

Quantity of substanceVSMeasurement precision

Solution Approach 1:

The patent applies the taking out principle by extracting and removing noise portions from the radar data sequence. The system identifies noise portions through frequency domain analysis and the trained auto-encoder model, then removes them before sequence restoration, thereby maintaining data volume while improving classification accuracy by eliminating harmful noise elements.

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The patent converts the harmful noise portions into beneficial information by using frequency domain transformation to identify and isolate noise characteristics. The noise, while harmful in the time domain, provides frequency signatures that the auto-encoder learns to recognize and selectively remove, thus converting a harmful factor into a useful feature for improvement.

Inventive Principle:

Principle #22Blessing in disguise (Convert harm into benefit)

3Adaptability or versatility

If traditional data augmentation is applied to radar datasets, then data diversity is improved, but effectiveness deteriorates due to frame dropping

Engineering Contradiction:

Improvedata diversityVSAvoidaugmentation effectiveness

Core Design Contradiction:

Adaptability or versatilityVSReliability

Solution Approach 1:

The patent replaces traditional mechanical data augmentation methods (such as simple copying or flipping) with a frequency-domain-based restoration system. By substituting the mechanical augmentation approach with FFT-based frequency analysis and auto-encoder-based intelligent restoration, the system achieves both data diversity and reliability, as the restoration process generates meaningful filled frames rather than arbitrary duplicates.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The patent applies parameter changes by transforming the data from time domain to frequency domain, processing different parameters (frequency components) that are not apparent in the original time-domain data. This parameter transformation enables effective data augmentation that accounts for frame dropping patterns, improving both diversity and reliability of the augmented dataset.

Inventive Principle:

Principle #35Parameter changes

Applied Scientific Principles

This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.

Function Achieved in This Case

Restores full input sequences, reduces noise, and enhances data augmentation, resulting in improved sequence classification and gesture recognition in radar systems.

Implementation Method 1

restoring time domain information for the input sequence by performing an inverse fast Fourier transformation on the frequency distribution

Methodology Applied

Scientific EffectFourier transformation:

Data Source

PatentUS12530878B2Sequence processing for a dataset with frame dropping

Publication Date: 2026.01.20 QUALCOMM INC

AI summary

A computer-implemented method for restoring a sequence for a dataset with frame dropping includes receiving an input sequence. A set of features is extracted from the input sequence. A frequency distribution is determined for the input sequence based on the extracted features. Time domain information for the sequence is restored and in turn, data for the input sequence is augmented based on the restored time domain information. Additionally, noise is removed from the input sequence.