Adaptive Reference Distribution for Quantized Signal Anomaly Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for proactive fault monitoring in computer systems using low-resolution A/D converters struggle to detect subtle anomalies due to quantization, leading to decreased sensitivity and robustness over time as the monitoring period increases.
Innovation Solution
A system that constructs a reference distribution for quantized signal values, updates the mean and variance of this distribution, and adjusts a deviant distribution to reduce the offset between them, enhancing sensitivity for anomaly detection using Sequential Probability Ratio Tests (SPRT).
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Duration of action of stationary object
If conventional moving histogram technique is used for long-term monitoring, then monitoring duration is extended, but detection sensitivity decreases due to built-up inertia in collected data
Solution Approach 1:
The patent applies dynamics by making the reference distribution adaptive over time. Instead of using a fixed reference distribution that accumulates inertia, the system dynamically updates the reference distribution parameters (mean and variance) based on recently collected data. This allows the monitoring system to maintain high sensitivity throughout extended monitoring periods by continuously adapting to current signal characteristics rather than being constrained by historical data inertia.
2Device complexity
If low-resolution 8-bit A/D converters are used in sensors, then device complexity is reduced, but measurement precision deteriorates due to signal quantization
Solution Approach 1:
The patent applies parameter changes by transforming the quantized signal values into a statistical domain. Instead of attempting to resolve the quantization limitation at the hardware level, the system changes the parameter representation from individual quantized values to distribution characteristics (mean, variance, and higher-order moments). This statistical transformation enables precise anomaly detection despite the coarse quantization from 8-bit converters, effectively overcoming the measurement precision limitation through parameter transformation.
3Reliability
If fixed offset between reference and deviant distributions is maintained, then system robustness is improved, but detection sensitivity decreases for subtle anomalies
Solution Approach 1:
The patent applies dynamics by making the offset between reference and deviant distributions adaptive rather than fixed. The system dynamically adjusts the deviant distribution parameters based on the current reference distribution characteristics and the specific anomaly detection requirements. This dynamic adjustment allows the system to maintain robustness against false alarms while simultaneously improving sensitivity to subtle anomalies by optimizing the offset according to current operating conditions and signal characteristics.
Data Source
AI summary
One embodiment of the present invention provides a system that facilitates detecting an anomaly in a signal, wherein the signal is sampled to produce a set of possible quantized signal values. During operation, the system constructs a “reference distribution” for an “occurrence frequency” of a specific quantized signal value from the set of possible quantized signal values. The system then obtains a “deviant distribution” associated with the reference distribution, wherein the deviant distribution has an offset from the reference distribution to indicate an anomaly in the signal. Next, in response to a new occurrence of the specific quantized signal value, the system updates a mean and a variance of the reference distribution for the specific quantized signal value. The system also adjusts the deviant distribution for the specific quantized signal value based on the updated mean and the updated variance of the reference distribution for the specific quantized signal value.