Let’s put it bluntly: When it comes to data quality in human behavior research, your job starts way before data collection – creating the proper physical framework for your biometric study is the first vital step toward obtaining good, meaningful data. Sounds reasonable? It definitely does … and yet, complications (or even failure) in studies along with poor data quality most often occur due to small mistakes that could have easily been avoided. Often they happen because researchers and staff just didn‘t know about the basic knacks and essential prerequisites for successful data collection.
Does that sound like you? Don‘t worry, you’ve found the right place – we’re here to hook you up with everything you need to know to create the optimal physical environment for your biometric study and drive high-quality results.
Last week we kicked off a series of blog posts dedicated to best practices for maximum data quality and let you in on 5 bullet-proof tips to collect strikingly good eye tracking data.
Today, we take a look at facial expression analysis and how to set up the best-possible framework for facial capture in controlled lab settings.
Curious? We bet you are. Follow our 3 surprisingly simple, yet most effective tips to boost data quality to the max for facial capture (to read up on facial expression analysis, click here for an in-depth look at the science behind it).
1. Camera position
Good news: For facial capture, an ordinary webcam as integrated in most laptops works just fine. No need to spend a fortune on a high-end device. Setting up the environment for facial detection really is no rocket science. In actual fact, there are only a few things to consider before you can get going with your study (the interpretation of the data is another story):
- Make sure the respondent is seated comfortably in front of the camera (read why seating posture is essential for data quality and how to provide optimal seating comfort during studies here).
- Speaking of camera position: In order to obtain best-possible results, position the camera as centrally as possible. If necessary, adjust the camera angle.
- Working with a large monitor can raise difficulties, especially if the respondent is seated too close to the screen – the camera simply might not be able to pick up facial landmarks necessary for automated detection. In these cases, rather increase definition or zoom in the screen content so the respondent can be comfortably seated further away from the monitor while still having clear, unrestrained view of the presented stimulus material.
In order for the camera to detect facial features properly, lighting is essential (as is the case for eye tracking). Although this might not exactly strike you as overly hard to achieve (just switch on the light, right?), there are a few pitfalls you should be aware of:
- Direct sunlight or overly bright illumination will wash out any contrast on the face and make facial detection nearly impossible.
- The same goes for dim lighting or no illumination at all as these light sources will cast dark shadows on the face – a no go for facial capture.
If possible, choose ambient light or bounce lighting (light bouncing off a reflective surface onto the respondent in order to achieve a softer lighting effect) as these forms of uniform illumination provide the best suitable conditions for facial detection.
3. Visibility of facial landmarks
For facial capture, the visibility of facial landmarks (key components of the face including eyebrows, eyes, nose, and mouth) is essential. Make sure your respondents don’t wear
- overly large, horn-rimmed glasses covering the eyebrows or sunglasses
- Longer, unruly beards (well-groomed three-day stubble should have no impact)
- facial jewelry such as multiple oral and eyebrow piercings (we recommend to check for each respondent individually and see if face detection is possible)
- head coverings (hats, caps, beanies etc.) as they are likely to cast shadows down the face
- hairstyles partially covering the face (side-swept bangs, for example)
Contact our experts at iMotions to learn more about the dos and don’ts of facial expression analysis and how to create a proper study environment that will push data quality to new heights.