FIELD: data processing.
SUBSTANCE: invention relates to digital processing of speech information and can be used in communication devices. For each position of the "sliding window", values of harmonic powers are calculated. For the position of the "sliding window" in which there is no signal, threshold values are calculated for the number of detected harmonics, power of harmonics and for average values of power of harmonics. For the positions of the "sliding window", where the presence of a speech signal is possible, for each detected harmonic, the samples of the mixture of the signal and interference are shifted by a value equal to half the period of the harmonic. Obtained readings are summed up with the initial ones. Using the obtained readings, values of harmonic powers are calculated. Based on the results of comparing the obtained values and values obtained during the primary analysis, they are recorded as noise-like or speech-like interference, or as a speech signal. Average power value and the number of detected harmonics are calculated. If these values exceed the corresponding thresholds, then the signal is considered to be a speech signal. If a speech signal was recorded for the previous position of the "sliding window", then the presence or absence of the speech signal is determined from the results of evaluating the change in the values of harmonic frequencies and their powers. If the speech signal ceases to exist and its duration does not exceed the maximum value, then this signal is considered to be a speech signal. Otherwise, this signal is considered to be interference.
EFFECT: high efficiency of selecting a speech signal in the presence of acoustic interference.
1 cl, 3 dwg, 3 tbl
