Knowledge Base

Manual Procedure Video and Camera Analysis

Aim

To analyze facial expressions from video, or live from camera images.

Prerequisites

• You do not carry out analysis with The Stimulus Presentation Tool. When you use the Stimulus Presentation Tool, the analysis starts when the participant starts a test.
• You created at least one participant. See Add Your Test Participants
• You created at least one video or camera analysis for every participant. See Add Analysis Input

Procedure

1. For Video analysis – Position the video to where you want to start the analysis.
2. Click the Start analysis button at the bottom of your screen. The track bar shows the progress of the analysis, the analysis speed (in fps), the running end time of the analysis, the end time of the video and whether the video has audio.
3. During analysis, check the Model quality bar in the Analysis Visualization window. It should look like the one below, the colored bar must cross the dashed line. You can adjust the model quality, see Minimum Model Quality. If the quality of your camera image is not good enough, the text Could not find a face or Could not classify the face (FaceReader cannot model the face) will appear in the Analysis Visualization window. Probably either the lighting of the test person's face or the position of the camera is not optimal. See Camera and Accessories how to improve your setup.
4. Click the Stop Analysis button to stop the analysis.
5. Video analysis – to analyze more episodes, move the video to a new position and click the Start analysis button again. The analyzed episodes are shown on the track bar.

IMPORTANT The data of all the analysis intervals are saved in one log file. The text Not Analyzed will appear in the log file for the time points that were not analyzed. This can result in very big files.

6. To export analysis results, choose File > Export and select to export the results of the analysis, participant, or the entire project. See Export Analysis Results

Notes

• By default, videos are processed frame-by-frame. To speed up the analysis, change the sample rate. See Video
• To carry out all video analyses at once, press the Start batch analysis button on the toolbar of the project explorer. Please see the tips on Improving FaceReader's performance which are especially important when using batch analysis.
• To redo an analysis, click the Clear results button and start the analysis again. Exported log files will not be deleted.
• When you run a camera analysis, a new video frame is analyzed when the analysis of the previous frame is finished. If detection is demanding, the number of frames being analyzed per second may be lower than the camera frame rate. In this case FaceReader skips a frame to keep up with the camera frame rate. After the analysis is finished, FaceReader interpolates the results in the skipped frames to the frame rate of the recorded video. See also Camera Frame Rate and Samples in FaceReader in Set Up Your Project
• When you run a video analysis, the average frame rate may be lower than that expected from the frame rate of the video file. For example 27 fps when the video frame rate is 30 fps and you set to analyze every frame. In that case video plays slower than 1x during analysis, but all frames are analyzed.

  See also:

  • Speed Up Analysis
  • Video in Settings
  • Sample Rate in Settings
• To analyze part of the video image, for example because your video contains multiple faces and you only want to analyze one of them, click the Select area of interest button. The Draw the selected area window opens. Drag a window around the face that you want to analyze. Either click the Apply button to start analyzing the selected face or the Reset selected area button to delete the selected area and draw a new one.
• You can analyze multiple faces in a video with the Advanced Research Module.
• To increase the speed and the accuracy of analysis, FaceReader uses information from previous frames to analyze the current frame. For instance, if in the previous frame FaceReader could correctly model the face, a re-detection is not required, because FaceReader already knows the location of the face. Information about previous frames is also used to smooth the output of FaceReader. These enhancements lead to much faster analysis, and it also allows the face modeling algorithm to better fit the face in the image. A side effect of this is that results depend on the start position used.
• If you re-analyze the video of a camera analysis and notice that the video is not analyzed.

Source: FaceReader 10 Reference Manual (Help), Noldus Information Technology