.png){kind=icon}
Return
A Lightweight Waveform Recognition Algorithm Based on Differential Features and Its FPGA Implementation
.png){kind=icon}
OA
AI
DOI：10.3390/electronics15174044.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Fast and accurate recognition of basic waveforms is important in embedded sensing, signal testing, and hardware measurement systems, especially under strict constraints on computation, memory, and power consumption. To address this problem, this paper proposes a lightweight waveform recognition method based on first-order and second-order differential features, together with decision tree and random forest classifiers. The proposed method transforms sampled waveforms into sign-based statistical features in the differential domain and performs classification through counting, comparison, and logical decision operations, thereby avoiding complex convolution, floating-point computation, and large-scale parameter storage. This paper first analyzes the separability of typical waveforms under differential statistics and studies the effect of noise on classification error. On this basis, a hierarchical recognition framework is constructed for five common waveform types, namely sine, square, triangle, rising sawtooth, and falling sawtooth. The models are trained on a PC platform and then implemented on an FPGA using Verilog. The experimental results show that the random forest achieves an accuracy of 99.7%, while the decision tree achieves 98.6%. On the xc7a35tfgg484-2 FPGA, both models consume zero DSP and zero BRAM resources. The random forest uses 1854 LUTs and 352 FFs, while the decision tree uses only 239 LUTs and 242 FFs. These results demonstrate that the proposed method achieves a favorable balance between recognition accuracy and hardware cost. The method is therefore well suited to real-time waveform recognition on resource-constrained FPGA platforms.
Keywords:
waveform recognition
FPGA
decision tree
random forest
Journal
IF:
2.6
Papers:
1.1W
Citations:
4.7W
Organization
Cited Papers
ShareSaveLow-Power FPGA Implementation of Convolution Neural Network Accelerator for Pulse Waveform Classification.png)
Algorithms
OAAI
AlgorithmsIF0
OAAIShareSaveA Satellite Full-Waveform Laser Decomposition Method for Forested Areas Based on Hidden Peak Detection and Adaptive Genetic OptimizationREMOTE SENSINGOAAI
REMOTE SENSINGIF4.1
OAAIShareSaveShareSaveShareSaveShareSave
View more