MEMS Microphone Glitch Prediction for Zero-Delay Gain Changes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional digital microphones face challenges in achieving high dynamic range due to power consumption constraints, leading to inadequate glitch removal and introduction of latencies in the digital signal processing path, which affects audio quality and requires significant device trimming and effort.
Innovation Solution
Implementing a machine learning-based glitch prediction and correction system in mixed analog and digital systems, specifically for digital MEMS microphones, using an AI-inspired predictor to predict and correct glitches without filtering delays, reducing power consumption, and eliminating the need for device trimming by training the model offline on glitch-representative signal information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional glitch removal techniques (filtering, look-up tables, interpolation) are used, then glitches are removed, but latency is introduced and device trimming is required
Solution Approach 1:
The system performs preliminary action by predicting glitches before they occur in the signal stream. The glitch predictor analyzes incoming samples and proactively identifies upcoming glitches based on patterns learned during training, allowing the system to prepare correction actions in advance without introducing latency to the critical signal path.
Solution Approach 2:
The invention extracts the glitch prediction function into a separate machine learning model that operates independently from the main audio signal processing chain. This extracted predictor runs on training data offline and only applies corrections when glitches are predicted, avoiding the need for continuous filtering or interpolation that would introduce latency.
2Reliability
If digital microphones use high dynamic range ADC or AGC amplifier, then dynamic range is improved, but power consumption increases
Solution Approach 1:
The system changes the operating parameters of the ADC by using oversampling at a lower resolution rather than using a high-resolution ADC. This parameter change allows the digital microphone to achieve high dynamic range through digital signal processing and glitch correction, avoiding the need for power-hungry high-resolution ADC hardware.
Solution Approach 2:
The invention replaces the mechanical/analog approach of using high-resolution ADC hardware with a digital/software-based approach. Instead of relying on physical ADC capabilities, the system uses machine learning algorithms to predict and correct quantization glitches, substituting computational methods for hardware solutions and reducing power consumption.
3Reliability
If conventional glitch removal filtering is applied, then glitches are reduced, but unacceptable latency is introduced in the digital signal processing path
Solution Approach 1:
The glitch predictor serves itself by using patterns learned from training data to autonomously identify and correct glitches in real-time. The machine learning model is self-sufficient, requiring no external filtering or interpolation operations, and operates without introducing latency to the critical audio signal path.
Data Source
AI summary
Disclosed embodiments provide glitch prediction based on machine learning algorithms in mixed analog and digital systems, particularly directed to digital microelectromechanical (MEMS) multipath acoustic sensors or microphones, which allow seamless, low latency gain changes without audible artifacts or interruptions in the audio output signal.