Non-uniform cosine modulated filter banks using meta-heuristic algorithms in CSD space.

Area of Science:

• Digital Signal Processing
• Filter Bank Design

Background:

• Non-uniform filter banks are crucial for efficient signal decomposition.
• Cosine Modulated Filter Banks (CMFB) offer a simplified design approach.
• Quantization of coefficients, especially using Canonic Signed Digit (CSD), can degrade filter bank performance.

Purpose of the Study:

• To present an efficient design methodology for non-uniform CMFB using CSD coefficients.
• To investigate the impact of CSD quantization on filter bank performance.
• To improve filter bank performance degraded by finite precision CSD coefficients.

Main Methods:

• Prototype filter design using window, weighted Chebyshev, and weighted constrained least square approximation methods.
• Quantization of filter coefficients to CSD format using a look-up table.
• Modification and application of meta-heuristic algorithms including Artificial Bee Colony, Gravitational Search, Harmony Search, and Genetic algorithms to optimize performance.

Main Results:

• The proposed CMFB design utilizes CSD coefficients for reduced implementation complexity.
• Meta-heuristic algorithms effectively mitigate performance degradation caused by CSD quantization.
• The optimized CMFB designs demonstrate lower implementation complexity, power consumption, and area requirements compared to conventional designs.

Conclusions:

• Efficient non-uniform CMFB can be designed using CSD coefficients.
• Modified meta-heuristic algorithms are effective in improving the performance of CSD-based filter banks.
• The proposed approach offers significant advantages in terms of hardware implementation efficiency for signal processing systems.