Frequency-Domain Streaming Pipeline for Low-Latency Signal Processing

Solution Overview

Problem

Existing technologies face inefficiencies and hardware burdens due to the need to transform pulse-code modulation (PCM) data into data for each frequency, which is not suitable for processing techniques like deep learning or GPU-based data processing.

Innovation Solution

An apparatus and method that directly utilize data for each frequency converted from an analog signal as stream data, involving signal conversion and preprocessing to bypass unnecessary frequency decompositions, using methods like short-time Fourier transform and signal filters.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If PCM data is transformed into data for each frequency using hardware, then data processing for deep learning and GPU-based processing becomes possible, but hardware complexity increases and time delay occurs

Engineering Contradiction:

Improvecompatibility with deep learning and GPU processingVSAvoidhardware complexity

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The frequency transformation is performed in advance at the data generation side before transmission, so that when the data reaches the receiving end, it is already in the required frequency-domain format for deep learning processing, eliminating the need for complex hardware transformation at the processing end

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent replaces hardware-based frequency transformation mechanisms with software-based processing that operates on pre-transformed data, substituting complex hardware operations with more efficient software algorithms that process frequency-domain data directly

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Adaptability or versatility

If PCM data is transformed into data for each frequency using hardware, then frequency-based processing becomes possible, but processing time delay increases

Engineering Contradiction:

Improvefrequency-based processing capabilityVSAvoidprocessing time delay

Core Design Contradiction:

Adaptability or versatilityVSLoss of time

Solution Approach 1:

The frequency transformation is performed in advance at the data generation side before transmission, so that when the data reaches the receiving end, it is already in the required frequency-domain format for deep learning processing, eliminating the need for complex hardware transformation at the processing end

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent skips the time-consuming hardware transformation step by pre-transforming the data in the frequency domain at the source, allowing the receiving end to directly process the transformed data without additional transformation time, thus rushing through the overall processing pipeline

Inventive Principle:

Principle #21Skipping (Rushing through)

3Adaptability or versatility

If PCM data is transformed into data for each frequency, then frequency-domain processing becomes possible, but hardware processing burden increases

Engineering Contradiction:

Improvefrequency-domain processing capabilityVSAvoidhardware processing efficiency

Core Design Contradiction:

Adaptability or versatilityVSProductivity

Solution Approach 1:

The frequency transformation is performed in advance at the data generation side before transmission, so that when the data reaches the receiving end, it is already in the required frequency-domain format for deep learning processing, eliminating the need for complex hardware transformation at the processing end

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The data is transformed into frequency-domain representation at the source where it is generated, making the data self-sufficient in its transformed form for downstream processing, so that the receiving end does not need to perform additional transformation operations

Inventive Principle:

Principle #25Self-service

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

Enhances signal control efficiency, reduces hardware and latency, and optimizes software processing time for data streaming in applications such as autonomous driving sensors, audio, and video.

Implementation Method 1

converting an input signal into a signal for each frequency; sampling a sensor signal generated by a sensor provided to measure an analog signal into a digital signal using an analog-to-digital converter; applying pulse-code modulation (PCM) to the digital signal

Methodology Applied

Scientific EffectPulse-code modulation: Phase Modulation

Implementation Method 2

allowing the digital signal to pass through a signal conversion filter which derives the signal for each frequency, by using the digital signal

Methodology Applied

Scientific EffectSignal filtering: Filter (electronic)

Implementation Method 3

performing a reference transform to decompose a signal to which the pulse-code modulation is applied to a component according to a predetermined frequency band; the reference transform may include short-time Fourier transform (STFT)

Methodology Applied

Scientific EffectFourier transform:

Data Source

PatentUS20260113223A1Apparatus and method for providing streaming by using data for each frequency

Publication Date: 2026.04.23 UNDERMILI INC

AI summary

An apparatus and a method for providing streaming by using data for each frequency are disclosed. The method for providing streaming by using data for each frequency according to an exemplary embodiment of the present disclosure includes the steps of: converting an input signal into a signal for each frequency; and transmitting target data in the form of a stream including the signal for each frequency to a main server for providing streaming for a client terminal.