Abnormal Sound Detection Using Frequency Emphasis Templates
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing abnormal sound detection devices struggle to detect slight abnormal sounds generated by target apparatuses immediately after the onset of abnormalities, making it difficult to identify such sounds effectively.
Innovation Solution
An abnormal sound detection device that utilizes an EQ unit to transform sound pressure signals into frequency characteristics, emphasizes frequencies associated with abnormal sounds using abnormal sound templates, and determines the presence of abnormal sounds through cosine similarity analysis, facilitating the detection of slight abnormalities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If sound pressure data is analyzed directly without frequency transformation, then the analysis process is simple, but slight abnormal sounds cannot be detected effectively
Solution Approach 1:
The patent replaces direct time-domain sound pressure analysis with frequency-domain analysis using FFT (Fast Fourier Transform). This substitution allows the system to detect slight abnormal sounds by transforming the analysis approach from mechanical/time-based to frequency-based, enabling detection of subtle frequency components that would be imperceptible in raw sound pressure data.
Solution Approach 2:
The patent introduces an EQ unit as an intermediary component between sound acquisition and abnormal sound determination. This EQ unit performs frequency transformation and emphasizes abnormal sound frequencies, acting as a mediator that bridges the gap between raw sound data and detectable abnormal patterns, thereby improving detection precision without requiring direct modification of the core detection logic.
2Measurement precision
If abnormal sound detection is performed on raw sound pressure signals, then the processing is straightforward, but detection precision for slight abnormalities is insufficient
Solution Approach 1:
The patent changes the parameter domain from time-domain sound pressure signals to frequency-domain characteristics. By performing FFT transformation, the system converts sound pressure data into frequency spectrum data, where abnormal sounds manifest as distinct frequency components. This parameter change makes slight abnormalities detectable through frequency analysis rather than amplitude analysis.
Solution Approach 2:
The patent applies local quality enhancement by emphasizing specific frequency bands associated with abnormal sounds. The EQ unit selectively amplifies frequency components corresponding to abnormal sound characteristics while suppressing normal operation frequencies, thereby improving the local signal-to-noise ratio for abnormal sound detection without processing the entire frequency spectrum uniformly.
3Measurement precision
If frequency emphasis is applied to detect slight abnormal sounds, then detection precision improves, but processing complexity increases
Solution Approach 1:
The patent applies preliminary action by pre-defining abnormal sound frequency characteristics and storing them as reference data. Before actual detection, the system prepares frequency emphasis parameters and abnormal sound templates, which are then used during runtime detection. This preliminary preparation simplifies the real-time processing by avoiding complex adaptive adjustments during detection.
Solution Approach 2:
The patent uses copying by creating frequency-emphasized copies of the original sound signal. The EQ unit generates a frequency-emphasized sound signal that replicates the original signal's structure but with enhanced abnormal frequency components. This copied signal can be analyzed separately without affecting the original data, simplifying the detection process while improving precision.
Data Source
Figure 1~2
Figure 3~4
Figure 5~6
AI summary
An abnormal sound detection device is configured to detect abnormal sound in operation sound of a target apparatus that is a target object for inspection. The device includes: a sound acquisition unit (10) that is configured to obtain the operation sound of the target apparatus as a sound pressure time signal that is a time signal indicative of a time variation in intensity of sound pressure signals; a signal transform unit (22) that is configured to transform the sound pressure time signal to a frequency characteristic indicative of intensity of the sound pressure signals at respective frequencies; an abnormal sound emphasis unit (23) that is configured to obtain, based on the frequency characteristic, an abnormal sound emphasized signal obtained by emphasizing intensity of a sound pressure signal at a frequency in association with the abnormal sound of the target apparatus among the sound pressure signals at the respective frequencies in the frequency characteristic; and an abnormal sound determination unit (25) that is configured to determine whether the abnormal sound is included in the operation sound of the target apparatus based on the abnormal sound emphasized signal.