FSR-Research-2025-vj

Shield:

This work is licensed under a Creative Commons Attribution 4.0 International License.

Abstract

This study investigates the impact of speech on the accuracy of non-invasive sleep metric measurements using a force-sensitive resistor (FSR) sensor placed under a mattress. In this work, a test setup using an FSR Model 408 was developed. Two algorithms for estimating heart rate (HR) and respiratory rate (RR) were developed for this. Their functionality was verified using a Polar H10 heart rate sensor as a reference. Three different speech patterns were chosen: normal talking, whispering, and a sustained "O" sound. Results show that speech, particularly low-frequency sounds, can significantly affect estimation accuracy, with the "O" sound causing the most significant deviations. The findings highlight the need for further research into mitigating speech-related artefacts in FSR-based sleep monitoring systems.

Data information

The data files are not named with the most descriptive names; hence, a table of the dataset and an explanation are given below. If you would like more details about how the data was gathered, please check the Research Paper.

| Name | Description | |-------------|----------------------------------------------------------------------------------------------------------------------------------------------------| | beda | person lying in bed without sound | | bedb | person lying in bed without sound (longer time duration than a) | | bedcrfilt | attempt at a CR filter, measurement turned out not useful for this research since the implementation of the CR filter and ADC was done incorrectly | | bednormal | normal speech | | bedwhisper | whispering speech | | bedosound | "O" sound | | sitting_a | test measurement where the FSR sensor was placed under the leg whilst sitting on a hard surface. |

For each dataset, a .txt and .csv file are present. The .txt files are from the custom FSR setup, and the .csv files are from the Polar H10.