| 摘要 |
<p>According to the invention, the method consists in: cyclically converting, by Fourier transform, digital signals resulting from analog/digital conversions of signals provided by two microphones a fixed distance apart and receiving the said sound signal into two series of discrete data, each discrete datum of the said series representing the energy and the phase of a frequency channel of the spectrum of the said sound signal; determining the dominant angle of arrival of the said sound signal from the phase differences between the discrete data corresponding to the same frequency channels of the said series, the said dominant angle of arrival corresponding to the angle of arrival of the said speech signal; obtaining an instantaneous spectrum of the said sound signal; updating a noise spectrum by comparing, for each frequency channel of the said instantaneous spectrum, of which the absolute value of the difference between the said dominant angle of arrival and the angle of arrival of the relevant frequency channel with a tolerance threshold value, the energies of the said frequency channels of the said noise spectrum being updated with the aid of the energies of the frequency channels of the said instantaneous spectrum, of which the absolute value of the difference between the said dominant angle of arrival and the angle of arrival of these frequency channels is greater than the said tolerance threshold value; subtracting the said updated noise spectrum from the said instantaneous spectrum to obtain a spectrum of the said speech signal.</p> |
