| 1. |
- Gaizo, Dario Del, et al.
(author)
-
Adaptive Pre-Processing for Neural Network Hardware Deployment
- 2023
-
In: Proceedings of the 2023 IEEE 33rd International Workshop on Machine Learning for Signal Processing, MLSP 2023. - : Institute of Electrical and Electronics Engineers (IEEE).
-
Conference paper (peer-reviewed)abstract
- Neural Networks are gaining popularity in the signal processing (SP) field. In radar SP, micro-Doppler based classification with Convolutional Neural Networks has shown state-of-the-art results. Still few works focus on top-down hardware deployment, and ID spectrum-based classification approaches with wide amplitude dynamic lose significant accuracy during quantization. In this paper, the problem is tackled with an adaptive pre-processing strategy that is able to mitigate accuracy loss upfront, learning the optimal representation of the data based on the classifier loss. To show this, a Drone vs. Unknown spectral classification scenario is presented.
|
|
| 2. |
- Gaizo, Dario Del, et al.
(author)
-
Hardware Deployable Radar Spectrum-based CNN classifier for Drone targets
- 2023
-
In: 20th European Radar Conference, EuRAD 2023. - : Institute of Electrical and Electronics Engineers Inc.. ; , s. 131-134
-
Conference paper (peer-reviewed)abstract
- It is well known that AI technology is spreading its application in all technical fields. Among all, radar field may take benefit from AI usage because its ability to recognize the environment by using small amount of domain expertise. Nowadays, the upcoming drone threats are stimulating the radar engineers to seek for advanced solutions for their classification in a congested scenario. As Deep Learning based drone classification has shown its first promising results, still few studies aim at hardware implementation of neural networks trained on frequency-amplitude 1D spectra. In this study, a lightweight CNN is trained on radar collected data prior to be quantized and evaluated on FPGA hardware.
|
|
