MM-LMF: A Low-Rank Multimodal Fusion Dangerous Driving Behavior Recognition Method Based on FMCW Signals

Round 1

Reviewer 1 Report

In this paper, the authors presented a method of multi-featured low-rank multimodal fusion based on the use of FMCW radar, which utilizes the complementarity between data modalities of different dimensions to recognize dangerous driving behavior. 

The importance of the problem addressed in the study is high. It is easy to follow the text. The paper is well formatted and structured.

A few questions raised to authors.

1. In the section “Experimental design and analysis” authors of the study say “the camera installed in front of the co-pilot position has the best effect, and the overall sitting posture image data of the front-seat driver and occupant in the car is obtained”. Does this type of camera installation overlaps or interfere the driver's view?
2. Authors evaluate their method performance, but there is no information about running time, RAM/ROM usage of the proposed method, method complexity.

Author Response

Please refer to the attachment

Author Response File: Author Response.pdf

Reviewer 2 Report

Please fix line 163

Please rewrite section 5.2.3, it's not clear and written in a way that reflects a tutorial than a journal paper.

Please rewrite the starting paragraph of section 5.2.6, it's not very clear.

Please provide a performance comparison with other works in this field.

Please explain how the works from 34 to 37 are related to this research. It is not very clear from the explanation "Therefore, the context mod- 146
eling of low-rank tensors can solve the situation of the high complexity of the model and 147
has end-to-end fusion features, which can show good performance in emotion recognition, 148
action recognition, and gesture recognit"

 

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

The changes look good to me.