Table 7 Paired t-test results comparing the proposed Hybrid Transformer-CNN model with baseline configurations (CNN-only, Transformer-only, and CNN + Transformer without fusion) over five independent runs.

From: Recognizing American Sign Language gestures efficiently and accurately using a hybrid transformer model

Model

t-statistic

p-value

CNN-only

118.29

< 0.0001

Transformer-only

223.54

< 0.0001

CNN + Transformer

200.05

< 0.0001

  1. All p-values are below 0.0001, confirming statistically significant performance improvements.
Back to article page