These are a number of systems and devices for heart beat measurement, which can be used to measure heart function or exercise volume, psychological barometer such as stress or relaxes. However, since the measurement of biological signals are not stable due to the unexpected several noises, the user is asked to firmly attach the sensor, for example electrode, and also to keep quiet even for the measurement of the heart beat rate, Fast Fourier Transform (FFT) is used as the traditional method of the measurement of heart beat rate, while the signals with unexpected noises in the measured signal are ignored.
In this study, we propose a novel method of heart beat tracking and develop a wearable device to visualize the heart beat, namely HOTARU ("firefly" in Japanese). The developed device allows users to freely move and play without attaching the sensor or electrode firmly. A wearable device is developed to indicate the heart beat in real time with different color of LED. The color changes according to the heart beat rate, and blinks in synchronization with the heart beat pulse. The developed system is able not only to track the heart beat but also to interpolate it from the noisy signals in real time. The heart beat is extracted from the original signal of the photoplethysmographic (PPG) sensor, which contains the noise delivered by body movement or other unexpected causes. In the proposed method, when the system is not able to determine the heart beat due to the sensor’s alignment or temporary no pulse, the heart beat is interpolated based on the past signal and liner prediction algorithm.
Traditionally, since the measurement of heart beat is focused on the heart beat pulse itself, but the tracking accuracy depends upon the environment. It is usually not stable because of the noise, and users are asked to rest during the measurement. The intervals of heart beat pulse are low frequency from about 0.5Hz to 2.0Hz and these intervals is assumed not rapidly change. But precise intervals are difficult to be recognized from the measured signals only by using peak detection due to the noises, which are usually impulse noise and look like heart beat pulse. Therefore, FFT is used as the analysis of the heart beat pulse. In this study, a modified peak detection algorithm is employed by combining Kalman filter to predict the intervals of heart beat, which is based on the linear prediction and uniform distribution function. Assuming that these intervals do not rapidly change, next heart beat interval can be estimated from transition of previous several intervals.
Potential applications include the tool for kids to promote social interaction. The user testing with several people is planned. We also will implement the sound feedback according to heart beat pulse for computer games and VR avatar.
This study was supported in part by the Global COE Program on "Cybernics: fusion of human, machine, and information systems.”