Convex Optimization-Based Filter Bank Design for Contact Lens Detection

被引：0

作者：

Madhe, Swati ^{[1
]}

Holambe, Raghunath ^{[2
]}

机构：

[1] Cummins Coll Engn, Pune 411052, Maharashtra, India

[2] SGGS Coll Engn & Technol, Nanded 431606, India

来源：

COMPUTING, COMMUNICATION AND SIGNAL PROCESSING, ICCASP 2018 | 2019年 / 810卷

关键词：

Filter bank; Convex optimization; Frequency band errors; TRIPLET;

D O I：

10.1007/978-981-13-1513-8_79

中图分类号：

TP18 [人工智能理论];

学科分类号：

081104 ; 0812 ; 0835 ; 1405 ;

摘要：

We have designed a novel convex optimization-based filter bank (FB), which minimizes the frequency band errors and optimizes time-frequency localization at the same time. The designed FB is regular and satisfies the constraint of perfect reconstruction (PR). In convex optimization, we have optimized quadratic constrained quadratic programs by transforming it into a semidefinite program. We have also compared the frequency band errors and time-frequency localization of proposed FB with existing FB. We have used this FB for designing a new contact lens detection (CLD) system. The IIITD database has been used for this purpose. The results have been expressed in terms of correct classification rate (CCR). The superiority of the designed FB has been shown by comparing the results with other existing CLD systems. The newly designed FB can also be effectively used for various signal processing applications.

引用

页码：781 / 790

页数：10

共 21 条

[1]

[Anonymous], INT C BIOM

[2] Structure and design of two-channel filter banks derived from a triplet of halfband filters [J].