| 摘要 |
This invention enables voice energy to be distinguished from energy due to other sources such as telephone call progress signals. A portion of the audio spectrum is divided into a high and into a low frequency band and the energy content of these bands is examined. The audio energy of a signal is passed through a set of filters and compared. A frequency boundary crossing of an input signal subsequent to filtering is used to signify the receipt of a voice signal. By operating in this manner, voice, whose predominant frequency switches in a random fashion from being greater than to less than a cutoff frequency, can be distinguished from other signals either below or above the cutoff frequencies of the filters. The detection of voice energy is not begun until a predetermined threshold of energy has been received for a specified period of time. This specified period of time need not be continuous after a ringback signal is received. Subsequent to the receipt of a ringback signal, the voice detection mechanism will proceed as long as the total time above the threshold exceeds the specified time. |
