Fig. 2: Online Short-Time Hilbert Transform (STHT) and Robust Zero-Crossing Conjugate (RZCC) event encoding. | Communications Engineering

Fig. 2: Online Short-Time Hilbert Transform (STHT) and Robust Zero-Crossing Conjugate (RZCC) event encoding.

From: Low-power Spiking Neural Network audio source localisation using a Hilbert Transform audio event encoding scheme

Fig. 2

a The STHT kernel (top) which estimates the quadrature component xQ of a signal, and frequency response of the kernel (bottom). b A noisy narrowband input signal (x; blue), with the quadrature component obtained from an infinite-time Hilbert transform (xQ; dashed) and the STHT-derived version (\({\hat{x}}_{Q}\); orange). Note the onset transient response of the filter before and around t < 2 ms. Corresponding up- and down-zero crossing encoding events estimated from x and \({\hat{x}}_{Q}\) are shown at top. c For a given signal (solid), zero crossings events (red) are estimated robustly by finding the peaks and troughs of the cumulative sum (dashed), within a window (ZC window).

Back to article page