Cross-correlation (crosscorrelationProc.m)

The IHC representations of the left and the right ear signals is used to compute the normalised CCF in the FFT domain for short time frames of cc_wSizeSec duration with a step size of cc_hSizeSec. The CCF is normalised by the auto-correlation sequence at lag zero. This normalised CCF is then evaluated for time lags within cc_maxDelaySec (e.g., [-1 ms, 1 ms]) and is thus a three-dimensional function of time frame, frequency channel and lag time. An overview of all CCF parameters is given in Table 32. Note that the choice of these parameters will influence the computation of the ITD and the IC processors, which are described in Interaural time differences (itdProc.m) and Interaural coherence (icProc.m), respectively.

Table 32 List of parameters related to 'crosscorrelation'.
Parameter Default Description
cc_wname 'hann' Window type
cc_wSizeSec 0.02 Window duration in s
cc_hSizeSec 0.01 Window step size in s
cc_maxDelaySec 0.0011 Maximum delay in s considered in CCF computation

The script DEMO_Crosscorrelation.m demonstrates the functionality of the CCF function and the resulting plots are shown in Fig. 37. The left panel shows the ear signals for a speech source that is located closer to the right ear. As result, the left ear signal is smaller in amplitude and is delayed in comparison to the right ear signal. The corresponding CCF is shown in the right panel for 32 auditory channels, where peaks are centred around positive time lags, indicating that the source is closer to the right ear. This is even more evident by looking at the SCCF, as shown in the bottom right panel.


Fig. 37 Left and right ear signals shown for one time frame of 20 ms duration (left panel) and the corresponding CCF (right panel). The SCCF summarises the CCF across all auditory channels (bottom right panel).