US 12,244,757 B2
Computer-implemented detection of anomalous telephone calls
Filip Hoste, Jabbeke (BE); Guy Van Der Meeren, Zwijnaarde (BE); Brecht Desplanques, Lichtervelde (BE); and Kris Demuynck, Lauwe (BE)
Assigned to CALLTIC NV, Zwijnaarde (BE)
Appl. No. 17/758,915
Filed by CALLTIC NV, Zwijnaarde (BE)
PCT Filed Jan. 22, 2021, PCT No. PCT/EP2021/051504
§ 371(c)(1), (2) Date Jul. 15, 2022,
PCT Pub. No. WO2021/148636, PCT Pub. Date Jul. 29, 2021.
Claims priority of application No. 2020/5043 (BE), filed on Jan. 23, 2020.
Prior Publication US 2023/0048264 A1, Feb. 16, 2023
Int. Cl. H04M 3/22 (2006.01); G10L 17/02 (2013.01); G10L 17/04 (2013.01)
CPC H04M 3/2281 (2013.01) [G10L 17/02 (2013.01); G10L 17/04 (2013.01); H04M 2203/6027 (2013.01)] 17 Claims
OG exemplary drawing
 
1. A computer-implemented method for detecting an anomalous telephone call, comprising the step of analyzing a telephone call associated with at least two user devices, wherein the telephone call is analyzed remote from the at least two user devices, wherein the analysis of the telephone call comprises the steps of:
a) obtaining a call audio stream of said telephone call;
b) deriving from the call audio stream a first set of multiple features;
c) converting the first set of features to an embedding vector of multiple features, wherein the embedding vector and the first set differ, wherein the first set of features is converted via a model to an embedding vector of multiple features, wherein the model is configured based on a training plurality of first sets of multiple features derived from a corresponding training plurality of audio streams, wherein the model is configured for representing in the embedding vector variability over the training plurality of first sets of multiple features, without particularization to detecting anomalous telephone calls; and
d) determining occurrence, or a probability of occurrence, of an anomalous telephone call based on the embedding vector.