ADAPTIVE SELF-TRAINED COMPUTER ENGINES WITH ASSOCIATED DATABASES AND METHODS OF USE THEREOF

• US 20180301143A1
• Filed: 04/13/2018
• Published: 10/18/2018
• Est. Priority Date: 04/03/2017
• Status: Active Grant

• Current Assignee: Green Key Technologies, Inc. (VoxSmart Limited)
• Original Assignee: Green Key Technologies LLC (VoxSmart Limited)
• Inventors: Shastry, Tejas, Tassone, Anthony, Kuca, Patrick, Vergun, Svyatoslav
• Granted Patent: US 11,114,088 B2
• Time in Patent Office: Days
• CPC Class Codes: G10L 15/02 Feature extraction for speech recognition; Selection of recognition unit
• CPC Class Codes: G10L 15/063 Training
• CPC Class Codes: G10L 15/08 Speech classification or search
• CPC Class Codes: G10L 15/187 Phonemic context, e.g. pronunciation rules, phonotactical constraints or phoneme n-grams
• CPC Class Codes: G10L 15/26 Speech to text systems G10L15/08 takes precedence
• CPC Class Codes: G10L 15/28 Constructional details of speech recognition systems
• CPC Class Codes: G10L 17/04 Training, enrolment or model building
• CPC Class Codes: G10L 2015/025 Phonemes, fenemes or fenones being the recognition units
• CPC Class Codes: G10L 2015/088 Word spotting
• CPC Class Codes: G10L 25/18 the extracted parameters being spectral information of each sub-band

Abstract

In some embodiments, the present invention provides for an exemplary computer system which includes at least the following components: an adaptive self-trained computer engine programmed, during a training stage, to electronically receive an initial speech audio data generated by a microphone of a computing device; dynamically segment the initial speech audio data and the corresponding initial text into a plurality of user phonemes; dynamically associate a plurality of first timestamps with the plurality of user-specific subject-specific phonemes; and, during a transcription stage, electronically receive to-be-transcribed speech audio data of at least one user; dynamically split the to-be transcribed speech audio data into a plurality of to-be-transcribed speech audio segments; dynamically assigning each timestamped to-be-transcribed speech audio segment to a particular core of the multi-core processor; and dynamically transcribing, in parallel, the plurality of timestamped to-be-transcribed speech audio segments based on the user-specific subject-specific speech training model.

Claims

1. A computer system, comprising:

at least one adaptive self-trained computer engine;

at least one multi-core processor comprising a plurality of cores, wherein the at least one multi-core processor is associated with the at least one adaptive self-trained computer engine so that the at least one multi-core processor is configured to receive executing instructions from the at least one adaptive self-trained computer engine;

wherein the computer system is configured to perform at least the following operations;

electronically receiving, by the adaptive self-trained computer engine, an initial speech audio data generated by a microphone of a computing device, wherein the initial speech audio data corresponds to at least one initial utterance received by the microphone from a particular user wherein the at least one initial utterance corresponds to an initial text being read by the particular user;

dynamically segmenting, by the adaptive self-trained computer engine, the initial speech audio data and the corresponding initial text into a plurality of user phonemes;

dynamically querying, by the adaptive self-trained computer engine, at least one database object containing at least one computer dictionary of stored subject-specific phonetic pronunciations of subject-specific words which are related to a particular activity area to match the plurality of user phonemes to a plurality of subject-specific phonetic pronunciations to form a plurality of user-specific subject-specific phonemes;

dynamically associating, by the adaptive self-trained computer engine, a plurality of first timestamps with the plurality of user-specific subject-specific phonemes, wherein each first timestamp corresponds to a time segment location of a respective matched user phoneme in the initial speech audio data so as to form at least one user-specific subject-specific data dictionary of timestamped user-specific subject-specific phonemes;

dynamically generating, by the adaptive self-trained computer engine, a plurality of user-specific subject-specific training instructions by utilizing the at least one user-specific subject-specific data dictionary of timestamped user-specific subject-specific phonemes as a user-specific subject-specific speech training model, wherein each user-specific subject-specific training instruction comprises a training text and a synthetized user-specific speech audio of the training text;

dynamically outputting, by the adaptive self-trained computer engine, the training text of each user-specific subject-specific training instruction to the computing device.