| CPC G06V 20/52 (2022.01) [G06V 10/764 (2022.01); G06V 10/82 (2022.01); G06V 20/41 (2022.01)] | 21 Claims |
|
1. A method for obtaining anomalous constructs from multimodal Binary Large Objects (BLOBs) of surveillance data using a cognitive abstraction module, the method comprising:
receiving the multimodal BLOBs of surveillance data concurrently, wherein the multimodal BLOBS of data surveillance data comprise a sequence of image frames, a sequence of audio segments, and a sequence of measurements over a first time interval and a second time interval;
generating meta-descriptors for each of the multimodal BLOBs of surveillance data obtained over the first and second time intervals, wherein the meta-descriptors comprise alphanumeric characters that characterize a corresponding multimodal BLOB of surveillance data;
generating a cumulative frequency distribution of all the generated meta-descriptors obtained over the second time interval;
generating meta-descriptor embeddings for each of the generated meta-descriptors; and
identifying, using an anomaly assertion model, multimodal anomalous constructs from all the generated meta-descriptor embeddings, and validating the multimodal anomalous constructs by asserting the multimodal anomalous constructs obtained over the first time interval against the cumulative frequency distribution of all the generated meta-descriptors obtained over the second time interval, wherein the second time interval comprises two or more multiples of the first time interval.
|