N-Path Spectral Decomposition for Low-Power Acoustic Classification

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current sound classification technologies face challenges in achieving accurate real-time classification at low power consumption and in an always-on fashion, particularly in Internet of Everything (IoE) systems, due to high computational demands and data latency associated with cloud processing.

Innovation Solution

The use of N-path filters for spectral decomposition and summary statistics in a mixed-signal design for feature extraction, which reduces energy consumption and allows for efficient classification without significant accuracy loss, implemented in an analog front-end followed by a digital back-end classifier, enabling lower computational and digitization energy costs.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If deep learning techniques are used for sound classification, then classification accuracy is improved, but power consumption and computational complexity increase significantly

Engineering Contradiction:

Improveclassification accuracyVSAvoidpower consumption

Core Design Contradiction:

Measurement precisionVSUse of energy by moving object

Solution Approach 1:

The patent segments the sound classification task into two parts: (1) spectral decomposition performed by analog N-path filters in the front-end, and (2) classification performed by a simplified digital back-end. This segmentation allows the energy-intensive spectral analysis to be done in the analog domain with lower power consumption, while the digital portion handles only the lighter classification task.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent replaces the digital computational system with an analog N-path filter system for spectral decomposition. By using analog circuitry to perform the mathematically intensive Fourier transform-like operations, the system avoids the high power consumption associated with digital signal processing, achieving comparable spectral analysis with significantly reduced energy usage.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Power

If cloud processing is used for sound classification, then computational power is sufficient, but data latency and privacy issues arise

Engineering Contradiction:

Improvecomputational powerVSAvoiddata latency

Core Design Contradiction:

PowerVSLoss of time

Solution Approach 1:

The patent enables the IoE system to perform sound classification independently using the analog N-path filter front-end combined with a simplified digital classifier. This self-service capability eliminates the need for cloud processing, allowing real-time classification to be performed locally on the device without introducing data latency or privacy concerns.

Inventive Principle:

Principle #25Self-service

3Use of energy by moving object

If analog N-path filters are used for spectral decomposition, then energy consumption is reduced, but implementation complexity increases

Engineering Contradiction:

Improveenergy consumptionVSAvoidimplementation complexity

Core Design Contradiction:

Use of energy by moving objectVSDevice complexity

Solution Approach 1:

The patent merges the spectral decomposition function with the filtering function by implementing N-path filters that perform both operations simultaneously. This integration eliminates the need for separate spectral analysis hardware, reducing overall system complexity while maintaining the energy efficiency benefits of analog processing.

Inventive Principle:

Principle #5Merging (Combining)

4Adaptability or versatility

If all-digital processing is used, then design flexibility is high, but combined energy of computation and digitization increases

Engineering Contradiction:

Improvedesign flexibilityVSAvoidcombined energy of computation and digitization

Core Design Contradiction:

Adaptability or versatilityVSUse of energy by moving object

Solution Approach 1:

The patent substitutes digital computation with analog N-path filter processing for spectral decomposition. This replacement eliminates the need for high-rate analog-to-digital conversion and subsequent digital Fourier transform computations, significantly reducing the combined energy of computation and digitization while preserving design flexibility through the programmable nature of the analog filters.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Data Source

PatentUS11763827B2N-path spectral decomposition in acoustic signals

Publication Date: 2023.09.19 THE BOARD OF TRUSTEES OF THE LELAND STANFORD JUNIOR UNIV

AI summary

A method and device for extracting information from acoustic signals receives acoustic signals by a microphone, processes them in an analog front-end circuit, converts the processed signals from the analog front-end circuit to digital signals by sampling at a rate of less than 1 kHz or more preferably less than 500 kHz; and processes the digital signals by a digital back-end classifier circuit. The analog front-end processing decomposes the received signals into frequency components using a bank of analog N-path bandpass filters having different subband center frequencies.