US 12,142,261 B2
Audio type detection
Krishna Khadloya, San Jose, CA (US); Chandan Gope, Cupertino, CA (US); and Vaidhi Nathan, San Jose, CA (US)
Assigned to Nice North America LLC, Carlsbad, CA (US)
Filed by INTELLIVISION TECHNOLOGIES CORP., San Jose, CA (US)
Filed on Mar. 16, 2021, as Appl. No. 17/203,269.
Application 17/203,269 is a continuation of application No. 16/280,806, filed on Feb. 20, 2019, granted, now 10,978,050.
Claims priority of provisional application 62/632,421, filed on Feb. 20, 2018.
Prior Publication US 2021/0210074 A1, Jul. 8, 2021
Int. Cl. G10L 15/16 (2006.01); G06F 9/54 (2006.01); G06N 3/045 (2023.01); G06N 3/08 (2023.01); G06N 20/10 (2019.01); G06N 20/20 (2019.01); G10L 25/18 (2013.01); G10L 25/21 (2013.01)
CPC G10L 15/16 (2013.01) [G06F 9/542 (2013.01); G06N 3/045 (2023.01); G06N 3/08 (2013.01); G06N 20/10 (2019.01); G06N 20/20 (2019.01); G10L 25/18 (2013.01); G10L 25/21 (2013.01)] 15 Claims
OG exemplary drawing
 
1. A system for event detection and classification, the system comprising:
a processor circuit; and
a processor-readable media comprising instructions that, when performed by the processor circuit, configure the processor circuit to:
receive audio information about an audio event;
apply a first audio event classification algorithm to determine whether the audio information includes an indication of a particular event, wherein the first audio event classification algorithm includes at least one of a support vector machine model, a logistic regression model, or a decision tree model, without utilizing a convolutional neural network;
identify a multi-dimensional spectrogram using the audio information; and
in response to determining, by the first audio event classification algorithm, that the audio information includes the indication of the particular event, apply information about the spectrogram at an input to a different second audio event classification algorithm that includes a convolutional neural network-based deep learning algorithm, wherein the different second audio event classification algorithm uses reference data trained target samples mixed with background noise, and wherein an output from the different second audio event classification algorithm includes an identification of the audio event as the particular event.