where m and j are the frame index and the index of frequency bands.
ω k = 2πk/N, k is the index of frequency bins, and N is the frame length.
This method not only describes the temporal output of the distributed power generation but also calculates reliability indices of frequency and duration, and is thus extensively applied.
The resulting power feature vector at frame with component index of frequency comprises the spectrotemporal power representation of the auditory response.
Finally, we limit the minimum value of β to 0.001, that is, the final β is obtained by: widehat{beta} k)= max left{beta (k),{beta}_{min}right} (34 where k is the index of frequency bins, β min = 0.001.
From the second to the last time frame, the difference of quantization index in each two adjacent time frames is calculated: diff_index tf,k) = index _q tf,k) - index _q tf-1,k), (15). in which tf is the index of time frame, t f>1, k is the index of frequency bins, k=0,…23, and i n d e x_q tf-1,k is the quantization index for azimuth of frequency bin k in time frame tf.
Given the frequency amplitude of two loudspeakers, X 1(k) and X 2 k), respectively, k is the index of frequency bin and ICLD can be calculated in each frequency Bark band b (b∈{0,1,2,..,23}) as: ICLD(b) = 10log_{10} left(frac{{E_{1} (b)}}{{E_{2} (b)}}right), (1).
In this way, the value of ( tilde{p} ) can be calculated by the following: tilde{p} k)= max left{ min left[mu cdot varXi left(b,kright)+upsilon, {tilde{p}}_{max}right],{tilde{p}}_{min}right} (28 where b denotes the index of the critical bands and k is the index of frequency bins.
The definition of LSD is given as: {d}_{mathrm{LSD}}=frac{1}{L}{displaystyle sum_{l=0}^{L-1}sqrt{frac{1}{N}{displaystyle sum_{k=0}^{N-1}{left[10{ log}_{10}frac{left|widehat{X}left l,kright)right|^2}{left|Xleft l,kright)right|^2}right]}^2}}} (36 where l is the frame index, k is the index of frequency bins, L is the total frames, and N is the frame length.
Second, since the speech energy usually decreases as frequency increases, for the calculation of p value at the intermediate frequencies, the linear decreasing of p is proposed as a function of the frequency, i.e., p k)={p}_{max }-frac{kleft({p}_{max }-{p}_{min}right)}{N} (27 where k is the index of frequency bins, N is the frame length, and p(k) denotes the p value of the kth frequency bin.
