Automatic Quality Assessment of Speech-Driven Synthesized Gestures

Figure 2

Human-likeness scores to every gesture synthesis systems. Here, the red line segment is the confidence intervals, Red line is the median value, yellow x represents outiler symbol, and yellow line in the rectangle is the mean value, and SA to SE are 5 different generation system, GT represents grand truth, BA and BT represent system with only audio and system with only text.