Scholarlink

<p>Even though face recognition is now an essential part of modern security, surveillance and authentication systems its large-scale adoption in real-time, limited resource environments poses persistent challenges. Although conventional convolution neural networks (CNNs) and Principal Component Analysis (PCA) methods are effective in controlled environments, their computational overhead limits their adoption for Internet of Things (IoT) applications and edge devices. Additionally, the ability to withstand occlusion and noise, adversarial attacks, and dynamic environmental variations is still not fully explored, while multimodal biometric fusion, which involves merging facial information with signature or fingerprint features, has yet to be explored. This study presents a superior face recognition algorithm that employs Multi-task Cascaded Convolution Networks (MTCNN) for face detection, FaceNet embeddings for feature extraction, and Vedic Mathematics’ computational accuracies, such as the Urdhva-Tiryak Sutra, Anurupyena Sutra, and Karatsuba multiplication. Additionally, The proposed Vedic Cosine Similarity eliminates the need for traditional floating-point operations and instead enables integer-based proportional scaling, vertical-crosswise multiplication, and fast multi-digit multiplication, which saves computational time and money. Despite the lack of recognition precision, experimental testing on real-time video streams and benchmark datasets shows significant improvements in processing speed throughput and energy efficiency. The enhancements facilitate dependable deployment on low-power, multi-camera surveillance networks, with reduced latency for real-time security monitoring. Its framework design is designed to meet emerging national security, border control and public safety needs as well as lay the foundation for future integration with quantum computing and multimodal biometric systems. Through this work, significant research gaps are addressed by enhancing computational efficiency and scalability while also providing greater robustness for real-time face recognition in mission-critical applications.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Harnessing ancient Vedic mathematics with Karatsuba multiplication for efficient and secure face recognition in surveillance

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年1月25日

查看全文
  • Debmalya Sanyal,
  • Gourab Saha,
  • Avijit Pakhira,
  • Abhrendu Bhattacharya,
  • Paramita Sarkar

摘要

Even though face recognition is now an essential part of modern security, surveillance and authentication systems its large-scale adoption in real-time, limited resource environments poses persistent challenges. Although conventional convolution neural networks (CNNs) and Principal Component Analysis (PCA) methods are effective in controlled environments, their computational overhead limits their adoption for Internet of Things (IoT) applications and edge devices. Additionally, the ability to withstand occlusion and noise, adversarial attacks, and dynamic environmental variations is still not fully explored, while multimodal biometric fusion, which involves merging facial information with signature or fingerprint features, has yet to be explored. This study presents a superior face recognition algorithm that employs Multi-task Cascaded Convolution Networks (MTCNN) for face detection, FaceNet embeddings for feature extraction, and Vedic Mathematics’ computational accuracies, such as the Urdhva-Tiryak Sutra, Anurupyena Sutra, and Karatsuba multiplication. Additionally, The proposed Vedic Cosine Similarity eliminates the need for traditional floating-point operations and instead enables integer-based proportional scaling, vertical-crosswise multiplication, and fast multi-digit multiplication, which saves computational time and money. Despite the lack of recognition precision, experimental testing on real-time video streams and benchmark datasets shows significant improvements in processing speed throughput and energy efficiency. The enhancements facilitate dependable deployment on low-power, multi-camera surveillance networks, with reduced latency for real-time security monitoring. Its framework design is designed to meet emerging national security, border control and public safety needs as well as lay the foundation for future integration with quantum computing and multimodal biometric systems. Through this work, significant research gaps are addressed by enhancing computational efficiency and scalability while also providing greater robustness for real-time face recognition in mission-critical applications.