A rapid prediction algorithm suitable to portable device for animal body-temperature measurement

Abstract

The body-temperature is the most significant vital signs of human and animals. It is easily imaginable that measurement of body-temperature of animals will be much more difficult than that of human being due to lack of endurance or fear of measuring instrument. Infrared temperature measurement device may be a solution, however coverage of hair and fur may incur a large error. To address this issue, a rapidly executed algorithm is developed for prediction of steady state body temperature, which needs only a few one-tenth of measurement duration that the currently popular machine learning-based approach usually requires. Let a cubic function $c(t)$ fit the sampled temperature data which are generated by the measurement within a significantly short duration from $t_{n-k}$ to $t_n,k>0$ . Then let a quadratic function $f(t)=a_2{t}^2+a_{2}t+a$ as a prediction function go through the point ( $t_n,c(t_n)$ ) and share the same slope of $s_n$ thereat. Finally try to find a next point ( $t_{n+m},f(t_{n+m})$ ), $m>0$ , where the slope satisfies $s_{n+m}=s_n/2$ and $m$ depends strongly on $s_n$ through an empirical formula. Accordingly, $f(t)$ can be determined by ( $t_n,c(t_n)$ ), $s_n$ and $s_{n+m}$ . Experiments indicate that the maximum of $f(t)$ approaches well the steady state temperature of the measured subject with a quite small error.