Skip to main content
. 2020 Sep 4;20(18):5030. doi: 10.3390/s20185030

Table 2.

Performance Comparison of proposed framework for driver identification.

Input Algorithm Accuracy FLOPs Memory Feature
Engineering
Windowing
60 × 45 DeepConvLSTM 97.72 1.624 M 7.88 MB Yes Wx = 60, dx = 6
60 × 45 DeepConvGRU 95.19 1.623 M 7.88 MB Yes Wx = 60, dx = 6
60 × 45 DeepConvLSTM-Attention 97.86 1.632 M 7.91 MB Yes Wx = 60, dx = 6
60 × 45 DeepConvGRU-Attention 98.36 1.631 M 7.91 MB Yes Wx = 60, dx = 6
60 × 15 FCM-LSTM 95.1 0.56 M 3.28 MB No Wx = 60, dx = 10
60 × 15 Proposed DepthConv-LSTM 97.78 0.235 M 1.74 MB No Wx = 60, dx = 10
60 × 15 Proposed DepthConv- GRU 98.52 0.234 M 1.74 MB No Wx = 60, dx = 10
40 × 15 Proposed DepthConv-LSTM 97.86 0.233 M 1.69 MB No Wx = 40, dx = 6
40 × 15 Proposed DepthConv- GRU 98.72 0.232 M 1.69 MB No Wx = 40, dx = 6