Telephonic communications – Call or terminal access alarm or control
Reexamination Certificate
1999-10-01
2003-10-14
Tsang, Fan (Department: 2645)
Telephonic communications
Call or terminal access alarm or control
C379S191000, C379S088010
Reexamination Certificate
active
06633632
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of Invention
This invention relates to a method and apparatus for determining the number of speakers on a call.
2. Description of Related Art
Currently, unscrupulous individuals use a variety of techniques to defraud network service providers. For example, one technique is for an individual to fraudulently set up a telephone call with no intention of paying for it and to leave the call up for extended periods of time which may exceed many hours or several days. The fraudster typically places the long distance call from, say, New York City to Hong Kong by charging the call to a stolen calling card number. Co-conspirators in the fraud, which are located in each city, work with the local population to “sell” time on the call. The call is then left up for a very long time while the locals take turns using the connection to speak to acquaintances in the distant city. Accordingly, there is a need for new processes and technology to detect and prevent such frauds from occurring.
SUMMARY OF THE INVENTION
The present invention provides a speaker counting device that is coupled to a network for monitoring calls placed over the network. The speaker counting device samples a call to determine the number of speakers participating in the call. Therefore, the speaker counting device allows the network service provider to automatically determine how many different speakers have participated in a call over some period of time.
It has been recognized that as the number of speakers which participate in a call increases, the probability that the call is a fraudulent one also increases. This is especially true if the same speakers do not consistently “reappear” in the voice traffic over the duration of the call.
The present invention determines the number of speakers on a telephone call by using automatic speech processing techniques, such as voice identification and voice verification. Such techniques are not required to depend on the language being spoken or the content of the speech. For example, the speaker counting device can sample speech and attribute that sample to a first voice record (Voice
1
). Later, the speaker counting device can again sample the call and test whether a detected voice is the same as Voice
1
. If the voice sample does not match a previously identified voice (e.g., Voice
1
), then the speaker counting device would attribute the sampled voice segment to a second voice record, Voice
2
. Again, the speaker counting device may sample a portion of the call and test whether Voice
2
is still speaking. If the voice sample does not match the previously identified voice record of Voice
2
, then the speaker counting device would then test whether the voice sample matches the previously identified voice record of Voice
1
. If the voice sample does not match Voice
1
, then the voice sample would be assigned to the next available voice record, Voice
3
. The speaker counting device would then repeat this process for as many speech samples as are taken during the call, and thereby count the number of different speakers participating in the call.
The more speech samples the speaker counting device uses, the more accurate the speaker counting device is in determining the number of speakers. The frequency of sampling and the length of the samples can be controlled in a variety of ways. In addition to fixed frequency or length, they could be random or controlled in more complex ways. For example, they can be database driven or dynamically controlled by a program according to the confidence level in identifying the speaker based on earlier speech samples. Furthermore, the sampling rate and/or length of the samples can be varied based on a growing suspicion of fraud. Additionally, as each voice sample is attributed to an existing voice record, the speaker counting device updates the corresponding voice data characteristics in a memory, thus making future voice identification more accurate.
Additionally, the speaker counting device can also determine long periods of silence which are another predictor of whether a call is fraudulent.
REFERENCES:
patent: 5623539 (1997-04-01), Bassenyemukasa et al.
patent: 6061648 (2000-05-01), Saito
patent: 6363145 (2002-03-01), Shaffer et al.
patent: 6442265 (2002-08-01), Harlow et al.
Harlow John Bruce
Sayko Robert
Wycherley Bruce D.
AT&T Corp.
Escalante Ovidio
Tsang Fan
LandOfFree
Method and apparatus for detecting the number of speakers on... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method and apparatus for detecting the number of speakers on..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method and apparatus for detecting the number of speakers on... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3140123