Dynamic Audio Artifact Prevention with Neural Network Resource Tuning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Audio processing systems experience delays due to sharing resources with other systems, leading to audio glitches like pops, crackles, and gaps, which degrade the user experience, and existing solutions either increase latency or power consumption to prevent underrun conditions.
Innovation Solution
A neural network-based system predicts the probability of audio underrun conditions and adjusts processing or memory frequencies and buffer sizes to prevent glitches dynamically, using deep neural networks to learn and optimize system parameters.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the audio buffer size is increased to prevent underrun conditions, then the reliability of audio output is improved, but the latency increases
Solution Approach 1:
The patent applies dynamics by making the audio buffer size adjustable rather than fixed. The system dynamically modifies buffer size based on real-time predictions of underrun probability from neural networks, allowing the buffer to adapt its size according to system conditions. This resolves the contradiction by enabling smaller buffers (lower latency) when underrun risk is low and larger buffers (higher reliability) when risk is high.
Solution Approach 2:
The patent changes the parameter of buffer size dynamically based on predicted underrun probability. By using neural network predictions to determine when to increase buffer size, the system can maintain reliability during high-risk periods while using smaller buffers during low-risk periods, thus balancing the latency-reliability tradeoff.
2Speed
If the processing frequency is increased to reduce audio latency, then the speed of audio processing is improved, but the power consumption increases
Solution Approach 1:
The system dynamically adjusts processing frequency based on neural network predictions of audio underrun probability. When underrun risk is detected, the system increases processing frequency to prevent glitches, accepting higher power consumption only when necessary. When risk is low, it operates at lower frequency to minimize energy usage, resolving the speed-power consumption contradiction.
Solution Approach 2:
The patent changes the processing frequency parameter dynamically rather than maintaining a constant high frequency. By using AI predictions to trigger frequency changes only when needed, the system achieves fast audio processing when required while minimizing overall power consumption during normal operation.
3Productivity
If the priority of audio processing threads is increased to reduce delays, then the productivity of audio processing is improved, but the system stability deteriorates due to increased resource contention
Solution Approach 1:
The system dynamically adjusts thread priority based on predicted underrun probability rather than maintaining a fixed high priority. When the neural network predicts low underrun risk, the system lowers audio thread priority to reduce resource contention and improve system stability. When risk is high, it temporarily raises priority to ensure productivity, thus balancing the contradiction.
Solution Approach 2:
The patent changes the priority parameter of audio processing threads dynamically. By using AI-based predictions to control priority changes, the system achieves high productivity when needed while maintaining better system stability during normal operation, avoiding the constant resource contention that would result from permanently high priority.
Data Source
AI summary
The disclosure is directed to a process that can predict and prevent an audio artifact from occurring. The process can monitor the systems, processes, and execution threads on a larger system/device, such as a mobile or in-vehicle device. Using a learning algorithm, such as deep neural network (DNN), the information collected can generate a prediction of whether an audio artifact is likely to occur. The process can use a second learning algorithm, which also can be a DNN, to generate recommended system adjustments that can attempt to prevent the audio glitch from occurring. The recommendations can be for various systems and components on the device, such as changing the processing system frequency, the memory frequency, and the audio buffer size. After the audio artifact has been prevented, the system adjustments can be reversed fully or in steps to return the system to its state prior to the system adjustments.