PPG (Photoplethysmography) is used to give the "blood volume pulse (BVP)" signal shown in the bottom graph (red signal on gray background) below.
The PPG is mainly used to identify the heart rate of the person wearing the sensor.
Heart rate is computed by detecting peaks (beats) from the PPG and computing the lengths of the intervals between adjacent beats. The inter-beat-interval (IBI) timing is used to estimate the instantaneous heart rate as well as to estimate average heart rate over multiple beats.
The main benefit of the Empatica's PPG is that it is more robust to motion than most other heart-rate sensors.
Empatica's PPG uses both green and red light. The green data contains the main information of the heart beats, while the red data contains information on the movements. By combing the two we are able to remove more motion artifacts that are not purely related to a global motion of the device but also to artifacts created by movements of tendons below the skin.
Keep the PPG data quality under control
The device must be worn reasonably snugly (but not uncomfortably tight) in order to be sure of getting good quality data. You can test the quality with the E4 realtime application. The picture below tells you what a good PPG signal looks like.
Example of good PPG data
Example of PPG data that is not highest quality.
Know how the data are obtained to understand the strengths
Empatica provides two types of data: the PPG data and the IBI data. They are strictly related. The sequence below describes the processing chain that Empatica employs
GREEN, RED -->[ Algorithm 1] --> PPG--> [Algorithm 2]--> IBI
Both PPG and IBI are publicly available in the Connect but Algorithm 1 is a property of Empatica and is not disclosed. Empatica has disclosed how Algorithm 2 operates in another article here.
The artifact removal process in the photoplethysmography is a double step process that can be managed both by Algorithm 1 and 2.
Different companies have their own approaches; for instance the Mio Alpha sensor uses an approach that takes as input the GREEN signal and the Accelerometer data.
GREEN,ACC -->[ Mio algorithm 1] --> PPG --> [Mio Algorithm 2] --> Average heart rate
How much is this different from Empatica's method? We compared the E3 and the Mio alpha in a separate article. A quick summary is:
- Mio Algorithm 1 is very good in a running scenario because the repetitive movements of the body are tracked in a good way by the accelerometers
- Mio Algorithm 2 averages the IBI values over a long period of time, making it less likely to spike with local movements that are not large repetitive movements. This is good for "average heart rate" over time but is not good for looking at variations in heart rate that are of interest to researchers looking at emotion or stress.
So what is the Empatica strategy ? We are more focused on scientific quality data than on high-level consumer questions such as "how high did my heart rate get during my run?" We thus want to get more fine-grained information that may be on the scale of seconds instead of minutes.
- The Empatica software aims to detect "every" heart beat. This make it useless in conditions with huge movements, where the Empatica Algorithm 2 will simply discard a lot of heart beats; however, when it is highly valuable when it detects good beats because it will produce IBI's that can be used to measure heart rate variability. Heart rate variability is of huge interest in studies of stress and its impact on medical conditions.
What should I expect from the IBI file?
This is a guide that will help you in deciding what to do depending on your experimental setting
- Studies in a static condition (i.e. pictures, movies, lessons) - use the IBI as provided.
- Studies where movements hold for less than 30% of the time - use the IBI as provided but be prepared to visually inspect the IBI's and use only segmented portions that are good. (There are also now starting to be algorithms to automate the detection of good regions).
- Studies with strong movement (more than 30% of time). For these, you will probably not be able to get enough reliable IBI's to compute heart rate variability continuously. However, you can still compute average heart rate like consumer devices have done for over a decade. For example you can use the motion information that our sensor gives, and when it is large, discard the IBI's with huge motion. Then average the remaining IBI's and check that they are in a reasonable range that has not changed too abruptly from the last estimate.
What works best will depend on what your specific goals are. Empatica research sensors give you access to the raw data so that you can make the best decisions for your application.
What other features does Empatica offer?
Empatica provides the full raw information on the distance between heart beats called IBI (Inter beat intervals).
At this time we do not offer any frequency domain features like the standard deviation of IBI or the Power spectral density at different bands. All those features could be relevant and belong to the so called Heart rate variability analysis. We expect to be offering more of these in the near future.
See this interesting article: Task Force of the European Society of Cardiology, and Task Force of the European Society of Cardiology. "the North American Society of Pacing and Electrophysiology. Heart rate variability: Standards of measurement, physiological interpretation, and clinical use." Circulation 93.5 (1996): 1043-1065.