Table 6.

Mathematical Description of the selected features for classification of LS signal $L\left[n\right]$.

Feature	Mathematical Representation
Standard Deviation (SD)	$SD=\sqrt{\frac{{\sum }_{i=1}^N{\left(Li-L_{Mean}\right)}^2}{n}}$ $where L:L\left[n\right]$
Peak to Peak (PP)	$PP=L_{max}-L_{min}$ Where $L_{max}$ and $L_{min}$ is the minimum and maximum value in the time domain
Log Energy (LE)	$LE=\log \left[{\sum }_{i=1}^N{\left(\left|Li\right|\right)}^2\right]$
Spectral Standard Deviation (SSD)	$\sigma =\sqrt{\frac{{\sum }_{i=1}^N{\left(Li-\left|L_{Mean}\right|\right)}^2}{n}}$ $where L:L\left[\omega \right]$
Spectral Skewness (SSkw)	$SSkew=\frac{2}{n}\left[\frac{{\sum }_{i=1}^{n/2-1}{\left(\left|L_i\right|-\left|L_{Mean}\right|\right)}^3}{{\sigma }^3}\right]$ $where L:L\left[\omega \right]$
Spectral Kurtosis(SK)	$SK=\frac{2}{n}\left[\frac{{\sum }_{i=1}^{n/2-1}{\left(\left|L_i\right|-\left|L_{Mean}\right|\right)}^2}{{\sigma }^4}\right]-3$ $where L:L\left[\omega \right]$
Spectral Flux (SF)	$SK=\frac{2}{i}\sqrt{{\sum }_{i=0}^{i/2-1}{\left(\left|L_{n,i}\right|-\left|L_{n, i-1}\right|\right)}^2}$
Spectral Roll Off (SRO)	If $m_{th}$ DFT coefficient corresponds to the spectral roll-off of the $k_{th}$ frame, then ${\sum }_{i=1}^n{L}_k\left(i\right)=C\sum _{i=1}^{F_L}{L}_k\left(i\right)$ C is the adapted percentage: 95% and $L: L\left[\omega \right]$
Spectral Decrease (SDec)	$SDec=\frac{{\sum }_{i=1}^{i/2-1}\frac{1}{i}{\left(\left|L_i\right|-\left|L_0\right|\right)}^{}}{{\sum }_{i=1}^{1/2-1}\left|L_i\right|}$ $where L:L\left[\omega \right]$
Mel frequency cepstral coefficient (MFCC)	In MFCC, (i) Frame blocking or windowing to get 50 to 60ms. (ii) Performing a discrete Fourier transform (iii) computing logarithm of the signal. (iv) Deforming the frequencies on a Mel scale, followed by applying the discrete cosine transform (DCT). Mel scale is calculated as follows: $Mel Scale=2595 lo{g}_{10}\left(1+\frac{f}{700}\right)$ ‘f’ refers to frequency ranges from 0 to fs.
Gammatone Frequency Cepstral Coefficient (GFCC)	In GCC, (i) Firstly, the signal is passed through gammatone filter bank which consists of 64 Channels. (ii) Take the absolute value at each channel and reduce it to 100 Hz as a way of time windowing. (iii) Take cubic root on the time-frequency representation. (iv) Deforming the frequencies on an equivalent rectangular bandwidth (ERB) scale Apply DCT to derive cepstral features. ERB scale is calculated as follows. $ERB=Alo{g}_{10} \left(1+hz\left(0.00437\right)\right)$ where $A = \frac{1000lo{g}_e\left(10\right)}{\left(24.7\right)\left(4.37\right)}$ ‘hz’ refers to frequency ranges i.e. 0-fs.