Basic steps of audio source localization

1. Bandpass filter input signals (300Hz~6KHz) of every channel
2. For first winSize samples, use as noise samples. (for compensation use)
3. Use Hann window to taper every winSize input signals
4. Find the peak value and its channel number
5. Use Hilbert Transform to that channel and find the peak location
6. shift back some samples and get the processing window which includes the peak signal.
7. Whiten the window signals
8. for every pixel, get the weight value for every channel and delay the signals of most channels
9. calculate the Steered Response Coherent Power for ever pixel. Coherent power differs from regular power which squared sum normalized by time. In coherent power, only the power from signal cross correlation between two channels is used in the summation. (x1[i]+…xN[i])^2 - (x1[i]^2+…xN[i]^2). where xn[i] is delayed signal of one channel. Large positive value of SRCP indicates a strong  correlation or coincidence between  signals in different channels.