APP Type Detection Using Kernel Signals for Frequency Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods fail to accurately determine the type of running applications on electronic devices, leading to incorrect adjustments in processing circuit frequency, which affects user experience.
Innovation Solution
An APP detection method using a K-nearest neighbors (KNN) algorithm to predict the type of running applications, collecting kernel information, and dynamically adjusting the processing circuit frequency based on the prediction result.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional APP detection methods are used, then the system can determine application types, but the accuracy is insufficient leading to incorrect frequency adjustments
Solution Approach 1:
The system pre-collects kernel information (CPU usage, memory usage, network activity, file operations) during APP execution before making frequency adjustment decisions. This preliminary data collection enables more accurate APP type classification through the KNN algorithm, resolving the contradiction by ensuring sufficient measurement precision before reliability-critical frequency adjustments are made
Solution Approach 2:
The system implements a feedback mechanism where the classified APP type information is used to adjust processing circuit frequency, and the system continuously monitors kernel information to validate and refine future classifications. This closed-loop feedback improves both detection accuracy and frequency adjustment correctness over time
2Productivity
If the processing circuit frequency is dynamically adjusted based on APP type, then device performance is improved, but the system complexity increases
Solution Approach 1:
The system introduces an intermediary classification layer (KNN algorithm) that sits between raw kernel information collection and frequency adjustment execution. This intermediary translates complex multi-dimensional kernel data into simplified APP type categories, enabling performance optimization without directly exposing the full complexity of the detection system to the frequency control mechanism
Solution Approach 2:
The system changes the parameter representation from raw kernel metrics to classified APP types. By transforming the data into categorical labels (e.g., gaming APP, benchmark APP, third-party APP), the system simplifies the control parameters for frequency adjustment while maintaining the ability to optimize device performance based on application characteristics
Data Source
AI summary
An application (APP) detection method includes: after an APP is launched, obtaining a process identifier (PID) corresponding to the APP from a first library; performing a calling operation upon an application programming interface (API) according to the PID; in response to the calling operation, collecting kernel information corresponding to the APP from a second library according to the PID; and performing a prediction operation by executing a K-nearest neighbors (KNN) algorithm according to the kernel information, to generate a prediction result, wherein the prediction result indicates a type of the APP.