Contents I. Introduction
Digital filters are gaining widespread attention for being superlatively fit in several applications. These applications include audio and image enhancement [1]–[2], noise attenuation [3], electrocardiogram (ECG) and electroencephalogram (EEG) compression etc. [4]. Digital filters are majorly classified into two types based on the length of their impulse response; finite impulse response (FIR) and infinite impulse response (IIR) filter. Between these filters, the IIR filter is designed with least order, but introduces nonlinear phase response and unstable behavior due to feedback, which displaces its pole-zero location [5]. In contrast to above, the linear phase FIR filters have no feedback, which makes them inherently stable and easily realizable [5], hence are widely used in various applications such as communications, seismology, multirate systems etc. [6]–[8]. FIR filters are mostly designed by using conventional methods and optimization based methods [9]. In conventional filter design methods, windowing and frequency sampling are used, between which windowing offers very limited design flexibility since there is no precise control over frequencies (passband cut-off frequency, stopband cut off frequency) and transition width, which limits its application and used to design prototype filters such as low-pass, high-pass and band-pass only [9]. Unlike above, frequency sampling technique is used to design filters with desired frequency response characteristics. However; the main drawback in frequency sampling is the presence of a finite error at un-sampled points in the spectrum [9]. Nowadays, different optimization based methods are promptly opted for designing FIR filter. These techniques involve calculation of the best value in a given set, which could be either the maximum or minimum value of a function, often known as the objective function or cost function [10], [21]. Recently, several nature-inspired meta-heuristic algorithms have also been introduced to solve multi-dimensional and complex problems in multi-parametric environment, which have become very popular in different fields of engineering for evaluation of a cost function due to their simple population-based strategy. In this context, genetic algorithm (GA), particle swarm optimization algorithm (PSOA), cuckoo search algorithm (CSA), bat optimization algorithm (BOA), cat swarm optimization algorithm (CSOA) etc. are frequently employed algorithms, which have emanated from the natural biological and physical processes [10]. In GA, bio-inspired operators such as mutation, crossover, and selection are exploited to solve different objectives [9]–[11]. The main advantage of GA is its ability to deal with complex problems, and can be easily implemented in parallel computers [12]. But it has few drawbacks of high computational speed and varying convergence rate [9]–[10]. Similarly, PSO is a population based stochastic optimization technique, inspired by the social behavior of swarms or animals such as bird flocks or fish schools to find their food [9]–[10], [13]. PSO is mostly used technique due to its simple, robust, and flexible nature that offers no overlapping and mutation in calculations [9]–[10], [13]. But the main disadvantage of PSO is that it easily gets trapped into local minima in high dimensional space, and thus caught in premature convergence [14]. In BA, the echolocation of bats with varying loudness and emission rates are exploited, and is better than other algorithms in terms of accuracy and efficiency [15].
