Wearable monitors are increasingly used to measure physical activity in research, and new tools and techniques are continually emerging [1]. Recent innovations have improved the cost, size, and technical capability of various monitors [2], but accuracy has not increased at a commensurate pace [3-6]. Thus, there is a need for further innovation. Successful innovation will likely entail novel measurement paradigms, rather than incremental improvements on current techniques [6]. One of the most promising and underexplored paradigms is to integrate data from multiple types of sensors, rather than the traditional use of only accelerometer sensors [7,8].

Photoplethysmography (PPG) is an optical technology that may have potential to enhance physical activity measurement when combined with established inertial sensors (accelerometers and gyroscopes) [9]. Although PPG was first described nearly 90 years ago, it has only recently gained a high level of visibility for physical activity assessment [10-12]. This growth is reflected in Figure 1, which shows the results of a Scopus search for documents addressing physical activity and PPG. Roughly half of the identified studies (183/385, 48%) were published in 2020 or later, and roughly three-quarters (285/385, 74%) were published in 2017 or later.

To date, most applications of PPG for physical activity assessment have involved consumer-grade smartwatches [13-16]. A wealth of developmental research has also been reported in the engineering literature [17-20], but commercial products for research have rarely incorporated PPG sensors and even more rarely given access to raw PPG data (ie, the data recorded by the sensor itself, without any preprocessing applied) [21]. This is beginning to change, and as it does, there is a need to raise awareness of PPG and its potential contribution to monitor-based physical activity assessment. In particular, awareness is needed for raw PPG data since it provides an avenue for device-agnostic measurement and iterative, open-source refinement, similar to the standard for inertial sensing [22].

In this viewpoint paper, we present raw PPG as a new frontier in monitor-based methodology. To do this, we first provide an overview of the fundamentals of PPG for physical activity assessment, after which we describe the importance and availability of raw PPG data, as well as specific applications where it holds the most potential. Throughout, we highlight ways that raw PPG data can synergize with raw data from inertial sensors to overcome long-standing challenges (eg, measurement during cycling).

To understand the revolutionary potential of raw PPG data for physical activity assessment, it is helpful to consider a similar revolution that has already taken place with accelerometer data [53-55]. Historically, accelerometer-based devices provided only proprietary “activity counts” as their output, which led to intermonitor differences and a lack of flexibility to innovate with new data processing techniques [56-58]. Over time, raw acceleration data became commonplace, opening doors for standardization and innovation in physical activity research [2]. An especially noticeable result was that many researchers began to focus on techniques that combined research-grade products with open-source tools for data processing and analysis, thereby promoting streamlined and coordinated progress in the field [59-64].

The potential parallels for PPG data are striking. To date, most research with PPG has relied on proprietary outputs from consumer-grade devices, which have been used to track heart rate, atrial fibrillation, blood pressure, and more [65-76]. Intermonitor differences and lack of flexibility are thus limitations of current standards for PPG, in much the same way as they once were for accelerometry. Furthermore, concerns have frequently been raised about unannounced algorithm and firmware updates that can make consumer-grade technology undesirable in certain research contexts [77-81]. The advent of raw PPG data therefore offers many of the same benefits that have already been derived from raw accelerometer data, especially when pairing research-grade devices with open-source resources. But the full potential of raw PPG for physical activity research cannot be realized unless the market provides devices that are scalable for use in large studies.

Existing research involving raw PPG data has generally involved small-scale devices (sometimes custom-made) [19,21,27,30,82], specialized tools for hospital use [83-86], or smartphone technology [87-89]. While these studies have shown strong proof-of-concept, they have only sometimes been oriented toward physical activity research, and the availability of suitable devices for large assessments remains an issue. The best-known research-grade devices are likely the E4 and EmbracePlus from Empatica Inc, the Shimmer3+ GSR from Shimmer Sensing, and the LEAP from ActiGraph LLC. Each device has strengths and limitations, with major points of comparison being cost, comfort, and access to raw data. The Shimmer3+ GSR is the most affordable option and provides access to fully raw data from both PPG and inertial sensors. However, a potential limitation is its reliance on physical components (eg, wired probes that wrap around the fingers) that may be unappealing or uncomfortable for some participants. The EmbracePlus is the most expensive device and is a replacement for the E4. It is designed like a standard smartwatch and is therefore very comfortable, but it does not provide truly raw PPG data (nor did the E4 [21]). Specifically, the EmbracePlus preprocesses PPG data using a proprietary algorithm that produces a blood volume pulse waveform, which resembles but does not replace raw data as it would appear in a direct recording from the photodetector. The LEAP device falls between the other 2 in terms of cost and comfort but does provide access to fully raw data from PPG and inertial sensors.

As new and upgraded devices continue to emerge and provide access to raw PPG data, a key objective will be to apply, extend, and standardize the techniques from earlier proof-of-concept studies for use in large-scale physical activity assessments for research. The following section outlines several specific areas in which there may be greatest warrant for these efforts.

The most obvious application of raw PPG for physical activity assessment is heart rate monitoring, where a notable contrast exists between the wrist-based optical approach and standard electrode-based approaches involving chest-worn monitors (eg, heart rate straps and Holter monitors). The latter tend to have greater accuracy than the former [90,91] and yet can also be uncomfortable to wear, especially over long periods [92]. Conversely, wrist-based PPG devices can be comfortably worn over long periods and yet have lower accuracy than chest-worn monitors. One implication is that raw PPG may encourage participant compliance in long assessment protocols (eg, lasting a week or more, which is common in physical activity research). This could be especially valuable for interventions that assess change over time, since responsiveness is generally a greater concern than accuracy in those contexts. Moreover, the diminished accuracy compared with chest-worn monitors may be less of an issue in cases where the key outcome is categorical intensity rather than continuous heart rate (eg, if assessing time in heart rate zones, where measurement error would be a concern only at the boundaries between zones, rather than across the spectrum of continuous heart rates). Nevertheless, there is a definite trade-off between accuracy and comfort, and neither chest-worn nor wrist-worn monitors are the optimal choice for every research question. This makes the accuracy-comfort trade-off a critical consideration when selecting a monitor for research. With continued innovation and refinement, the accuracy gap may narrow between chest- and wrist-worn devices, and trade-off–related considerations may change accordingly. But it is unlikely that the issue will ever disappear completely.

While heart rate monitoring is an obvious application for raw PPG, it may not be the most impactful one. Rather, there may be greater promise when combining raw data from PPG and inertial sensors to predict other physical activity–related outcomes such as energy expenditure. This multimodal approach not only allows for robust correction of motion artifact in the PPG signal (as described previously) but also enables concurrent analysis of movement and cardiovascular data. While this is not an entirely new concept, the ability to carry it out using purely raw data from a single wrist-worn device is quite recent. Thus far, the primary multimodal methods have relied on either separately worn movement and cardiovascular monitors [93-95] or the chest-worn Actiheart device (CamNtech Ltd) [96-99] when predicting energy expenditure. These approaches have shown clear synergy between movement and cardiovascular data but have ultimately had limited uptake compared with the widespread use of wrist-worn monitors in field-based research. Furthermore, heart rate has been the only cardiovascular parameter emphasized with the earlier methods, whereas raw PPG can potentially lead to enhanced predictions through the capture of additional aspects of cardiovascular response to activity (eg, pulse wave parameters and variability). This highlights the warrant for translating and extending earlier concepts of multimodal assessment to the use of raw PPG and inertial data from wrist-worn devices.

The combination of raw PPG and inertial data may also help overcome known limitations of movement-only techniques in the prediction of energy expenditure. For example, wrist-worn monitors are generally unable to register any motion during cycling despite the level of lower-limb exertion, resulting in poor measurement validity [100]. In contrast, PPG may still detect exertion during cycling because it relies on optical and physiological signal rather than inertial signal. This advantage reflects the known benefit of using not only multiple sensors but multiple types of sensors [51,56]. Similar benefits may also arise for other activities where the body’s inertial profile is altered, such as when carrying an external load or pushing a stroller [101,102]. The potential to overcome these limitations with virtually no change in participant burden highlights the strong potential of raw PPG to improve physical activity research, especially in the area of energy expenditure prediction.

Combining raw PPG and inertial data may also be beneficial for activity recognition, promoting greater understanding about certain elements of activity context [103]. Activity recognition is also important as a precursor to energy expenditure prediction, since it is much easier to predict energy expenditure if the type of activity is first known [104-106]. This is the basis for several prior models of energy expenditure, including well-known 2-regression models [107-114]. The utility of raw PPG for activity recognition was recently highlighted by Hnoohom et al [115], who calibrated models using data from 3 open-access datasets [116-118]. The parent studies used devices from Shimmer, Empatica, and Maxim Integrated (Analog Devices Inc), and the models were calibrated using deep learning and different combinations of accelerometer, PPG, and electrocardiographic data. When combining accelerometer and PPG data, the dataset-specific models each achieved near-perfect accuracy during 10-fold cross-validation. However, external validations are needed to confirm the effectiveness of the models and apply them with scalable devices, as described previously.

Outside of energy expenditure and activity recognition, raw PPG may have use for measuring a range of other physical activity–related variables as well, from specific hemodynamic parameters related to exertion (eg, pulse transit time [119]) to consequences of physical activity such as fatigue and recovery [120]. Steps warrant mention as well, given their status as a well-known output of most wearable activity monitors. Raw PPG could potentially enhance or refine the measurement of steps and related variables (eg, cadence), including during periods where the inertial movement profile is altered, as described previously for energy expenditure. The combination of raw data from PPG and inertial sensors could also enable automated measurement of highly specialized outcomes in free-living settings, such as cardiac-locomotor coupling (ie, synchrony of footfalls with systole or diastole) [121].

Finally, although our focus has been on physical activity–related applications of raw PPG, it is important to acknowledge potential contributions in the broader context of 24-hour measurement as well. This refers to the growing emphasis on interrelationships between physical activity, sedentary behavior, and sleep as part of a daily composite [122,123]. The importance of the concept is reflected in the release of 24-hour movement guidelines from numerous governments and the World Health Organization over the last few years [124-129], and there are at least 2 key contributions raw PPG can make to 24-hour assessments. One is to differentiate between nonwear, sedentary behavior, and sleep, all of which produce minimal accelerometer and gyroscope signal and are therefore hard to tell apart using only inertial data. Raw PPG may exhibit richer variation across the categories, thereby assisting with disambiguation. Some proof-of-concept already exists in this area as well, given the amount of prior work using PPG for sleep measurement [130]. The other benefit may be to assist with classifying posture (seated or lying vs upright) [131], which is essential for differentiating sedentary behavior from light-intensity physical activity [132]. These possibilities highlight the strong potential and exceptional flexibility of raw PPG, which will be an asset for a broad range of movement-oriented research in the coming years.

In this viewpoint paper, we have introduced raw PPG and highlighted its potential benefits for physical activity assessments. Our specific focus on raw PPG data (as opposed to preprocessed or aggregated data) was critical and timely, given the recent emergence of mainstream devices that provide access to them. A key strength of raw PPG is that its optical basis complements the inertial basis of familiar accelerometer and gyroscope sensors. Furthermore, raw PPG can be used to assess not only heart rate but also broader aspects of cardiovascular physiology. These are the driving forces behind the potential we laid out in the prior sections.

Going forward, it will be critical to obtain raw data from both PPG and inertial sensors, not only to facilitate merging them but also to make new algorithms both transparent and device agnostic (ie, applicable to data from any PPG-inclusive device). These characteristics help combat the “black box” phenomenon of closed-source devices [133]. Device agnosticism also plays an important role in “future proofing” new methods by reducing dependence on individual monitors that can leave the market at any time (eg, as seen with the SenseWear Armband, Phillips Actiwatch, Empatica E4, and ActiGraph GT9X). The use of raw PPG can be further advanced by using open-source channels when developing and disseminating of new resources, consistent with growing standards for existing wearable devices in physical activity assessment [59-64].

This viewpoint paper was among the first to suggest the value of integrating raw PPG data into large-scale assessments of physical activity, where the importance of raw accelerometer and gyroscope data has already been established. In doing so, the viewpoint paper serves to orient new users to the wealth of prior work on PPG from other research areas, where critical reference points have been provided that can spur a paradigm shift in physical activity research. A noteworthy limitation of the viewpoint paper was that it was not a systematic review. As such, it did not fully summarize the available literature, whether in general or focused specifically on physical activity assessment. Nevertheless, our overall conclusion is that there is warrant for vigorous exploration of raw PPG going forward.