Estimating the cardiac signals of chimpanzees using a digital camera: validation and application of a novel non-invasive method for primate research

Abstract

Cardiac measures such as heart rate measurements are important indicators of both physiological and psychological states. However, despite their extraordinary potential, their use is restricted in comparative psychology because traditionally cardiac measures involved the attachment of sensors to the participant's body, which, in the case of undomesticated animals such as nonhuman primates, is usually only possible during anesthesia or after extensive training. Here, we validate and apply a camera-based system that enables contact-free detection of animals' heart rates. The system automatically detects and estimates the cardiac signals from cyclic change in the hue of the facial area of a chimpanzee. In Study 1, we recorded the heart rate of chimpanzees using the new technology, while simultaneously measuring heart rate using classic PPG (photoplethysmography) finger sensors. We found that both methods were in good agreement. In Study 2, we applied our new method to measure chimpanzees' heart rate in response to seeing different types of video scenes (groupmates in an agonistic interaction, conspecific strangers feeding, nature videos, etc.). Heart rates changed during video presentation, depending on the video content: Agonistic interactions and conspecific strangers feeding lead to accelerated heart rate relative to baseline, indicating increased emotional arousal. Nature videos lead to decelerated heart rate relative to baseline, indicating a relaxing effect or heightened attention caused by these stimuli. Our results show that the new contact-free technology can reliably assess the heart rate of unrestrained chimpanzees, and most likely other primates. Furthermore, our technique opens up new avenues of research within comparative psychology and facilitates the health management of captive individuals.

Keywords: Comparative psychology; Contact-free; Heart rate; Signal processing.

Fig. 1

Setup and data collection

Fig. 2

System overview

Fig. 3

Reflections on the window

Fig. 4

An example of CEEMDAN decomposition of the Y-channel signal in the detected ROI

Fig. 5

The frequency spectrum of decomposed IMFs (intrinsic mode functions)

Fig. 6

a Signals from a video processed without stabilization. b Signals from a video with using MOSSE Tracker to stabilize target. c Plot of raw signals from a and b

Fig. 7

Data comparison of Hope for the same time segment (550–560 s). a Raw data of the PPG finger sensor. b Smoothed PPG signal and its frequency spectrum. c The cardiac signal extracted from video and the frequency spectrum of the signal from video. d A noise part of the PPG finger sensor

Fig. 8

Heart rate values for Sessions 1, 3, and 9, respectively. The heartbeat estimated from the video signal is depicted in blue, and the red color depicts the heartbeat measured by the finger sensor. The yellow dash line boxes show the obviously unstable part of PPG sensor data

Fig. 9

Histograms of four pairs of the beat distributions

Fig. 10

Example images of the three types of videos shown to the chimpanzees. In the aggression condition, familiar conspecifics were engaged in an aggressive interaction. In the eating condition, unfamiliar individuals consumed large amounts of food. In the nature condition, natural habitats were filmed from the bird’s perspective

Fig. 11

Structure of a trial. Each video presentation was preceded and followed by the presentation of a grey screen during which we measured the chimpanzees’ baseline heart rates

Fig. 12

Heart rate change during video presentation relative to the pre-baseline in the three conditions. Dots represent individual data per session for each chimpanzee. Dashed line represents pre-baseline level. Turquoise stars depict average changes over all individuals and sessions with bootstrapped 95% confidence intervals