Sound Data Processing for Privacy-Safe Abnormality Detection

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing systems for detecting abnormalities in apparatuses via sound analysis often transmit human voice components, potentially infringing on privacy and failing to effectively distinguish abnormal sounds from normal operations.

Innovation Solution

An information processing apparatus that acquires sound intensity data, processes it to extract maximum values at a temporal resolution where human voice is unrecognizable, and transmits this data to an external apparatus for analysis, ensuring privacy protection and accurate abnormal sound detection.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If sound data is transmitted to external apparatus for abnormal sound analysis, then abnormal sound detection capability is improved, but human voice privacy is compromised

Engineering Contradiction:

Improveabnormal sound detection capabilityVSAvoidprivacy infringement

Core Design Contradiction:

Measurement precisionVSObject-affected harmful factors

Solution Approach 1:

The patent extracts and removes human voice components from the sound signal before transmission. The sound processing unit identifies and eliminates frequency bands characteristic of human speech, retaining only the mechanical sound components necessary for abnormal sound detection, thereby preventing privacy infringement while maintaining detection capability

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The patent applies frequency domain transformation and filtering to change the parameters of the sound signal. By converting time-domain sound data to frequency-domain representation and applying selective filtering, the system transforms the signal to remove voice components while preserving abnormal sound characteristics

Inventive Principle:

Principle #35Parameter changes

2Object-affected harmful factors

If sound signal is filtered to remove human voice components, then privacy protection is improved, but information useful for abnormal sound analysis may be lost

Engineering Contradiction:

Improveprivacy protectionVSAvoidabnormal sound analysis information

Core Design Contradiction:

Object-affected harmful factorsVSLoss of information

Solution Approach 1:

The patent applies different processing characteristics to different frequency bands of the sound signal. Frequency bands corresponding to human voice are heavily filtered or removed, while frequency bands containing mechanical abnormal sound characteristics are preserved with minimal processing, ensuring each frequency region receives appropriate treatment for its specific purpose

Inventive Principle:

Principle #3Local quality

3Object-affected harmful factors

If maximum value extraction is performed at coarse temporal resolution, then human voice recognizability is reduced, but temporal precision of sound events is degraded

Engineering Contradiction:

Improvehuman voice recognizabilityVSAvoidtemporal precision of sound events

Core Design Contradiction:

Object-affected harmful factorsVSMeasurement precision

Solution Approach 1:

The patent divides the sound signal into multiple time sections and extracts maximum values from each section. By segmenting the continuous signal into discrete intervals and selecting representative maximum values, the system reduces temporal resolution to prevent voice recognition while maintaining sufficient detail for abnormal sound detection through the distribution pattern of these maximum values

Inventive Principle:

Principle #1Segmentation

Data Source

PatentUS20240105214A1Information processing apparatus, non-transitory computer readable medium, and information processing method

Publication Date: 2024.03.28 FUJIFILM BUSINESS INNOVATION CORP

AI summary

An information processing apparatus includes a processor configured to: acquire first data indicative of a temporal change of an intensity of sound emitted by an apparatus; generate second data by extracting, from the first data, a maximum value in each section of a time width corresponding to temporal resolution at which human voice is unrecognizable and discarding values other than the maximum value; and transmit the second data to an external apparatus.