| 摘要 |
In an audio analyser of audio signals associated with an audio scene (eg. a soundtrack from multiple sources in fig. 2 for synchronisation), a spectral flatness value (eg. geometric mean of power spectrum divided by arithmetic mean such that 1 is pure noise) is determined and compared to a threshold. If above the threshold, the signal contains useful audio for time alignment. If the spectral flatness value is less than the threshold, the alignment signal generator (219 fig. 5) may instead output a detectable audio signal (eg. ultrasound) 455 for synchronisation (223 fig. 3). GPS or Camera Pose Estimates may be used to estimate the distance separating the sources in order to account for microphone sensitivity. |
