HMM-Based Adaptive Spectrogram Tracking for Heart Rate Monitoring
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current heart rate monitoring techniques using optical sensors, such as pulse oximeters, face challenges due to interference from white noise and motion artifacts, leading to unreliable and inaccurate heart rate measurements.
Innovation Solution
The implementation of a heart rate monitoring system that utilizes a Hidden Markov Model (HMM) to analyze spectrogram data from optical sensors, considering multiple possible heart rate change paths and adaptively adjusting parameters to filter out noise and accurately track heart rate changes, thereby improving accuracy and robustness.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If adaptive filters are used in preprocessing to reduce noise, then noise reduction is improved, but complete noise removal is not achieved and the system fails
Solution Approach 1:
The patent introduces an intermediary statistical model (Gaussian distribution parameters) between the raw spectrogram and the final heart rate detection. Instead of directly filtering the spectrogram, the system models the noise characteristics using mean and variance parameters, then uses these statistical intermediaries to guide the detection process through likelihood calculations, achieving more robust noise handling than direct filtering alone.
Solution Approach 2:
The system dynamically changes parameters (mean and variance of the Gaussian distribution) based on the observed spectrogram data. By adaptively adjusting these statistical parameters to match the current noise conditions, the system can optimally distinguish signal from noise in varying environments, rather than using fixed filtering parameters.
2Measurement precision
If the system searches for frequency components with maximum power, then detection sensitivity is improved, but dramatic and unreal jumping occurs in heart rate output
Solution Approach 1:
The patent performs preliminary actions by pre-calculating and storing the Gaussian distribution parameters (mean and variance) from the spectrogram data before the actual heart rate detection. This preliminary statistical characterization of the signal allows the subsequent detection to be more stable, as it references pre-computed noise characteristics rather than reacting to raw power fluctuations alone.
Solution Approach 2:
The system uses feedback by comparing the observed spectrogram against the modeled Gaussian distribution and using the likelihood calculation to guide heart rate detection. The detection process continuously references the statistical model parameters, creating a feedback mechanism that stabilizes output by considering the probability distribution of expected signals rather than relying solely on instantaneous power maximums.
3Measurement precision
If multiple possible heart rate change paths are tracked, then accuracy is improved, but computational complexity increases
Solution Approach 1:
The patent applies partial action by tracking multiple candidate paths through the spectrogram data, but not exhaustively searching all possible paths. Instead, it focuses computational effort on the most probable paths based on the Gaussian model and likelihood calculations, achieving high accuracy without the full computational burden of exhaustive multi-path tracking.
Data Source
AI summary
Embodiments of a system and method for heart rate monitoring are generally described herein. A method may include receiving, at a device, a spectrogram including a plurality of bins of represented frequencies corresponding to potential heart rate measurements, selecting a bin of frequencies from the plurality of bins of represented frequencies with an average power that exceeds average powers of remaining bins of frequencies of the plurality of bins of represented frequencies, determining whether a frequency of the bin of frequencies represents a valid heart rate based on the average power and a past valid heart rate, and in response to determining that the frequency represents a valid heart rate, outputting a heart rate corresponding to the frequency.