Depending what you are trying to measure and what is your target group. General consensus is accelerometers are the most accurate, but it can be costly with large sample sizes (e.g. 100+ participants). Actigraph is the most popular, has been validated in different populations, and is reliable for assessing intensity, duration, steps, sedentary time, etc. However, it won't pick up on activities like cycling or swimming (most are not waterproof).
You could always use this or validated surveys, and match either method with in-built accelerometers in smartphones (Apple has an in-built health app that tracks activity, and there are plenty of Android free apps to count steps/activity. Make sure you have a validation paper to back up your choices).
If you don't have budget constrains, a FitBit or Garmin would be great. Not many validation studies, but you will increase compliance for sure ( accelerometers are usually worn around the waist and participants might stop wearing because they forget to put it on or not very fashionable to wear)
If you use accelerometers, make certain you do an intra-unit comparison. They all record motion but they don't all record it in a completely similar fashion even within the same brand of monitor. Look at the research of RG McMurray.
We are in the middle of conducting a Systematic Review of the reliability and validity of consumer wearables and preliminary results would suggest that Actigraph is widely used and validated. A few things to consider though:
- if you are hoping to increase participant compliance with a brand-name device (eg. Fitbit, Garmin, Jawbone etc.) make sure to consider that the validity/reliability will change depending on the gait speed (some devices are more valid during jogging/running while others are more valid during walking)
- the reliability and validity of the wearables will vary depending on what variable you are considering (eg. step count, energy expenditure, HR etc.)
I'm working with accelerometers and we are working on a project to compare reliability and validity of them and some devices as watches. I suggest you to consider what kind of physical activity you want to measure and if you need to identify them (e.g. walk, run, or bicycling). If you want to measure steps and intensities of PA, the accelerometers could be useful. However, if you want to identify the kind of PA you couldn't do it by accelerometers.
In term of results, they could be influenced by the place where participants have to wear the accelerometers or another electronic devices. Besides, there are differences linked to cut off for each intensities of PA and wearing time.