DETAILED DESCRIPTION The above and other problems in the prior art are overcome and a technical advance is achieved in accordance with the present invention which relates to a DTMF detector installed in parallel with a voice coder running at the Internet telephony transmitting gateway.
In accordance with the invention, a very fast algorithm is used (e.
g.
, 10 milliseconds detecting time) to determine whether or not a tone is suspected in the data stream being transmitted.
If such a tone is suspected, then transmission is held up for several frames until a tone is confirmed.
If, prior to the expiration of the several frame period, a "kill" signal is received, then it is determined that the audio stream did not include a DTMF.
If the foregoing decision indicates that a DTMF tone is present, the actual digitized audio reflecting the DTMF signal is discarded, and a digital signal is sent, out of band, indicating to a receiving gateway that the DTMF tone has been depressed.
If, on the other hand, the algorithm at the transmitting gateway indicates that a DTMF tone has not been depressed, and that the suspected DTMF tone was actually a small portion of the audio signal that was simulating a DTMF tone, then the audio signal suspected of being a tone is transmitted.
Any delay introduced to the audio signal by the suspecting of tone and the later finding out that the tone was not actually present in the audio signal is compensated for by deleting later arriving portions of the audio signal which are determined to contain silence.
In another embodiment, the DTMF detector uses as its input the LPC filter present in the voice coder.
This has two advantages.
First, since the LPC encoding is being done as part of normal speech encoding, no additional overhead is introduced.
Second, the FFT utilized to detect the DTMF frequencies operates on the LPC coefficients, rather than on raw digital data.
Therefore, the DTMF tones can be detected with a relatively short FFT.
|
Multiplexer-related 
