Method and apparatus for identifying and relaying audio segments of interest in a monitored security environment

1 . An audio analysis device, the audio analysis device

comprising:

at least one processor; and

memory comprising a plurality of instructions that, when executed by the at least one processor, are configured to cause the at least one processor to:

receive administrator criteria that specifies an alarm event handling response;

obtain an audio file associated with an alarm event at a monitored security environment;

detect, based on a machine learning model, an audio anomaly in the audio file;

in response to detecting the audio anomaly:

generate an audio segment based on the audio anomaly, the audio segment comprising the audio anomaly;

generate metadata based on a first machine learning (ML) model and by assigning, based on the administrator criteria, a respective priority weight to each of a plurality of characteristics associated with the audio segment, at least one of the plurality of characteristics comprising:

a description of the audio anomaly;

a secret phrase;

a description of at least one other sound in the audio segment;

a location where the audio file was recorded;

an identity of at least one person present at the location;

a number of distinct speakers having voices present in the audio segment;

a distance between a source of the audio anomaly and the audio recording device; or

an identity of at least one background noise in the audio segment;

determine that the administrator criteria specifies providing the audio segment and the metadata in an agent portal for a monitoring agent assigned to the alarm event;

select, based on the administrator criteria and the respective priority weight of each of the plurality of characteristics, a subset of the metadata for rendering in the agent portal; and

encode the audio segment and the subset of the metadata for rendering in the agent portal.

2 . The audio analysis device of claim 1 , wherein the plurality of instructions is further configured to cause the at least one processor to:

select, based on the metadata, a second ML model for analyzing the audio segment; and

update the metadata based on the analysis by the second ML model.

3 . An audio analysis device, the audio analysis device comprising:

at least one processor; and

memory comprising a plurality of instructions that, when executed by the at least one processor, are configured to cause the at least one processor to:

receive administrator criteria that specifies an alarm event handling response;

obtain an audio file associated with a monitored security environment;

detect an audio anomaly in the audio file;

in response to detecting the audio anomaly:

generate an audio segment based on the audio anomaly, the audio segment comprising the audio anomaly;

generate metadata based on a first machine learning (ML) model and by assigning, based on the administrator criteria, a respective priority weight to each of a plurality of characteristics associated with the audio segment;

determine that the administrator criteria specifies providing the audio segment and the metadata in an agent portal;

select, based on the administrator criteria and the respective priority weight of each of the plurality of characteristics, a subset of the metadata for rendering in the agent portal; and

encode the audio segment and the subset of the metadata for rendering in the agent portal.

4 . The audio analysis device of claim 3 , wherein the plurality of instructions are further configured to cause the at least one processor to detect the audio anomaly in the audio file using the first ML model.

5 . The audio analysis device of claim 3 , wherein the plurality of instructions is further configured to cause the at least one processor to:

receive resolution information relating to the audio anomaly; and

train, based on the resolution information, the first ML model.

6 . The audio analysis device of claim 3 , wherein the plurality of instructions is further configured to cause the at least one processor to:

receive at least one change to the first ML model;

perform, based on the at least one change to the first ML model and on determining not to transmit, a further analysis of the audio file; and

based on the further analysis, one of transmit the audio file to an agent device or mark the audio file for deletion.

7 . The audio analysis device of claim 3 , wherein at least one of the characteristics comprises at least one of:

a description of the audio anomaly;

a secret phrase;

a description of at least one other sound in the audio segment;

a location where the audio file was recorded;

an identity of at least one person present at the location;

a number of distinct speakers having voices present in the audio segment;

a distance between a source of the audio anomaly and the audio recording device; or

an identity of at least one background noise in the audio segment.

8 . The audio analysis device of claim 3 , wherein the plurality of instructions are further configured to cause the at least one processor to:

select, based on the metadata, a second ML model for analyzing the audio segment; and

update the metadata based on the analysis by the second ML model.

9 . The audio analysis device of claim 3 , wherein the plurality of instructions are further configured to cause the at least one processor to select a subset of the metadata by at least selecting individual ones of the characteristics that have a respective priority weight above a priority threshold, the priority threshold being based on the administrator criteria.

10 . The audio analysis device of claim 3 , wherein the plurality of instructions is further configured to cause the at least one processor to increase, based on the metadata, a length of the audio segment.

11 . A method performed by an audio analysis device, the method comprising:

receiving administrator criteria that specifies an alarm event handling response;

obtaining an audio file associated with a monitored security environment;

detecting an audio anomaly in the audio file;

in response to detecting the audio anomaly;

generating an audio segment based on the audio anomaly, the audio segment comprising the audio anomaly;

generating metadata, based on a first machine learning (ML) model and by assigning, based on the administrator criteria, a respective priority weight to each of a plurality of characteristics associated with the audio segment;

determining that the administrator criteria specifies providing the audio segment and the metadata in an agent portal;

selecting, based on the administrator criteria and the respective priority weight of each of the plurality of characteristics, a subset of the metadata for rendering in the agent portal; and

encoding the audio segment and the subset of the metadata for rendering in the agent portal.

12 . The method of claim 11 , further comprising detecting the audio anomaly in the audio file using the first ML model.

13 . The method of claim 11 , further comprising:

receiving resolution information relating to the audio anomaly; and

training, based on the resolution information, the first ML model.

14 . The method of claim 11 , further comprising:

receiving at least one change to the first ML model;

performing, based on the at least one change to the first ML model and on determining not to transmit, a further analysis of the audio file; and

based on the further analysis, one of transmitting the audio file to an agent device or marking the audio file for deletion.

15 . The method of claim 11 , wherein at least one of the characteristics comprises at least one of:

a description of the audio anomaly;

a secret phrase;

a description of at least one other sound in the audio segment;

a location where the audio file was recorded;

an identity of at least one person present at the location;

a number of distinct speakers having voices present in the audio segment;

a distance between a source of the audio anomaly and the audio analysis device; or

an identity of at least one background noise in the audio segment.

16 . The method of claim 11 , further comprising:

selecting, based on the metadata, a second ML model for analyzing the audio segment; and

updating the metadata based on the analysis by the second ML model.

17 . The method of claim 16 , further comprising selecting a subset of the metadata by at least selecting individual ones of the characteristics that have a respective priority weight above a priority threshold, the priority threshold being based on the administrator criteria.

18 . The method of claim 11 , further comprising increasing, based on the metadata, a length of the audio segment.