| CPC G10L 25/78 (2013.01) [G06F 40/289 (2020.01); G06F 40/30 (2020.01); G10L 15/26 (2013.01)] | 8 Claims |
|
1. A method comprising:
receiving audio data captured by one or more microphones of a public safety portable radio at a crime scene, wherein the audio data comprises a first statement uttered by a first person and a second statement uttered by a second person;
detecting, via a speaker-diarization software module executed by one or more processors upon the audio data, that the first person uttered the first statement and the second person subsequently uttered the second statement;
detecting, using an NLP processing module, that the first statement includes an utterance indicative of predetermined criminal elements known to be associated with a crime;
capturing a voiceprint data of the first statement, in response to the utterance indicative of predetermined criminal elements;
detecting, using NLP processing, that the second statement is a Miranda statement, thereby identifying the second person as an officer;
receiving additional audio data captured by the one or more microphones after the Miranda statement is made by the officer, wherein the additional audio data comprises additional remainder speech content;
capturing a voiceprint data of the remainder speech content;
comparing the captured voiceprint data of the remainder speech content to the voiceprint data associated with the first statement made by the first person to determine a voiceprint match;
in response to the voiceprint match, determining that the first person is a suspect and that the remainder speech content is a post-Miranda statement uttered by the suspect;
generating, via a voice-transcription software module executed by the one or more processors upon the audio data, a transcription of the pre-Miranda statement, the Miranda statement, and post-Miranda statement, the post-Miranda statement containing the remainder speech content that was uttered by the suspect after the Miranda-statement was uttered by the officer;
detecting, via the NLP software module executed by the one or more processors upon the transcription, that a degree of semantic similarity between the pre-Miranda statement and the post-Miranda statement does not satisfy an admissibility condition for the crime;
presenting, via a user interface software module, controlling a user interface electronic display of the public safety portable radio, an indication that the suspect did not repeat the pre-Miranda statement in the post-Miranda statement;
generating, via the NLP software module executed by the one or more processors upon the transcription, a list of transcribed pre-Miranda statements identifying the elements of the crime that need to be repeated by the suspect and displaying the list on the user interface electronic display of the public safety portable radio;
generating, via the NLP software module executed by the one or more processors upon the transcription, a question to induce the suspect to repeat the pre-Miranda statement containing the criminal elements associated with the crime and listed on the list;
presenting the question via the user interface software module, to the user interface electronic display of the public safety portable radio; and
presenting the question via the user interface software module, to the user interface electronic display.
|