Integration in CNN and FIR filters for improved computational efficiency in signal processing
This research paper explains the design process of the 8 × 8 Vedic multipliers based on the “UrdhvaTiryagbhyam” Sutra in combination with the “Nikhilam Sutra“ and the Karatsuba algorithm. To effectively generate a 16-bit product, the used architecture consists of four four-by-four Vedic modules, an...
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2025-01-01
|
| Series: | Ain Shams Engineering Journal |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S2090447924005823 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | This research paper explains the design process of the 8 × 8 Vedic multipliers based on the “UrdhvaTiryagbhyam” Sutra in combination with the “Nikhilam Sutra“ and the Karatsuba algorithm. To effectively generate a 16-bit product, the used architecture consists of four four-by-four Vedic modules, an 8:1 carry-save adder, and two nine-bit binary adders. The UrdhvaTiryagbhyam approach splits multiplications into pieces, the Nikhilam Sutra uses the concept of binary complements, and the Karatsuba algorithm offers improvements in large numbers of multiplications. The proposed addition microarchitecture, which consists of using a Fast Carry Switching Adder and the Kogge-Stone Adder with associated selection signals and speculative logic, improves carry propagation time. The ability of the Vedic multiplier is tested within an FIR filter and a CNN processing element, revealing significant enhancements in speed and efficiency. Importantly, the proposed multiplier based on the modification of Vedic Nikhilam yields the lowest power consumption (248.93 mW), the lowest delay (27.95 ns), and the lowest PDP (6.96 pJ), thus making it appropriate for usage in HPC related to signal processing and neural network computations. Moreover, the developed FIR filter for the CNN and the EEG signal datasets were employed to detect seizures and Alzheimer’s disease. The incorporation of the Vedic multiplier into the CNN framework reveals the application of the proposed idea in the field of biomedical signal processing with improved computational speed and accuracy. The results corroborate the multiplier’s efficiency in decreasing the computational complexity and enhancing the possibility of real-time analysis of CNN-based systems in healthcare. |
|---|---|
| ISSN: | 2090-4479 |