Mattress Sensor-Based Respiration Rate Estimation Using Unsupervised Clustering

Abstract

This paper presents a novel approach for measuring the human respiration rate utilizing a mattress sensor. The proposed method employs an unsupervised clustering technique in conjunction with a cluster selection algorithm to accurately measure the respiration rate. The pressure exerted on the mattress sensor reflects the volume changes in the upper body during the breathing process, enabling the estimation of the respiration waveform based on the measured pressure variations. Through the clustering method, the measured pressure changes are effectively separated into respiratory-related clusters and noise. Subsequently, the cluster selection algorithm identifies the optimal combination of clusters that best represents the respiration pattern. To evaluate the performance of the proposed method, it was compared against two alternative methods, namely RWD and RAC. The results demonstrate that the proposed method, referred to as RCS, achieved the highest mean Pearson correlation coefficient of 0.88, indicating superior performance compared to the other methods. In contrast, the mean Pearson correlation coefficients for RWD and RAC were determined as 0.61 and 0.76, respectively. Statistical analysis confirmed the significance of the differences among these methods (p < 0.001). Furthermore, a regression analysis was conducted to assess the accuracy of the respiration rate measurements. The findings revealed that the proposed method yielded the most precise estimation of the respiratory rate.