# US 11,443,747 B2 · Artificial intelligence apparatus and method for recognizing speech of user in consideration of word usage frequency > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 11,443,747 B2 - Document ID: US11443747B2 - Status: Granted - Kind: B2 - Application number: 16/653,120 - Filed: 2019-10-15 - Published: 2020-02-06 - Granted: 2022-09-13 - Earliest claimed: 2019-09-18 - Assignees: LG Electronics Inc. - Applicants: LG Electronics Inc. - Inventors: Jonghoon Chae - Web page: https://patentgrep.com/patent/US11443747B2 ## CPC - G06F 40/20: Natural language analysis (semantic analysis of natural language G06F40/30) - G10L 15/26: Speech to text systems (G10L15/08 takes precedence) - G06F 17/18: for evaluating statistical data {, e.g. average values, frequency distributions, probability functions, regression analysis (forecasting specially adapted for a specific administrative, business or logistic context G06Q10/04)} - G10L 15/065: Adaptation - G10L 15/193: Formal grammars, e.g. finite state automata, context free grammars or word networks - G10L 15/22: Procedures used during a speech recognition process, e.g. man-machine dialogue - G10L 2015/223: {Execution procedure of a spoken command} - G10L 2015/226: {using non-speech characteristics} - G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N) - G10L: SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING ## Abstract Disclosed herein is an artificial intelligence apparatus for recognizing speech of a user including a microphone and a processor configured to obtain, via the microphone, speech data including speech of a user, determine a frequency weight for each word using a speech recognition log, generate a speech recognition result corresponding to the speech data using the frequency weight, and perform control corresponding to the speech recognition result. ## Claims ### Claim 1 An artificial intelligence apparatus for recognizing speech of a user, comprising: a microphone; and a processor configured to: obtain, via the microphone, speech data including speech of a user, for each of a plurality of sections of the speech data: calculate a probability for each of a plurality of words, determine, using a speech recognition log, a frequency weight for each of the plurality of words, for each of the plurality of words, calibrate the probability by multiplying the calculated probability by the determined frequency weight, convert the speech data into text by (i) selecting, for each section of the plurality of sections, a word having a highest calibrated probability and (ii) combining the selected words, and perform control based on the text. ### Claim 2 (depends on 1) The artificial intelligence apparatus of claim 1, wherein the processor is configured to: normalize the calibrated probabilities, and convert the speech data into the text by selecting, for each section of the plurality of sections, a word having a highest normalized calibrated probability. ### Claim 3 (depends on 1) The artificial intelligence apparatus of claim 1, wherein the processor is configured to: check a usage frequency of a word using the speech recognition log, and set, based on usage frequencies of words, a frequency weight of a word to be higher than a frequency weight of another word. ### Claim 4 (depends on 3) The artificial intelligence apparatus of claim 3, wherein the processor is configured to: determine a private frequency weight for a word using a private speech recognition log, determine a public frequency weight for a word using a public speech recognition log, and calibrate a probability for a word using the private frequency weight and the public frequency weight, wherein the private speech recognition log includes speech recognition logs in the artificial intelligence apparatus, and wherein the public speech recognition log includes speech recognition logs in an external apparatus. ### Claim 5 (depends on 4) The artificial intelligence apparatus of claim 4, wherein the processor is configured to: determine an integrated frequency weight by calculating a weighted sum of a private frequency weight of a word and a public frequency weight of a word, and calibrate a probability using the integrated frequency weight. ### Claim 6 (depends on 1) The artificial intelligence apparatus of claim 1, wherein the processor is configured to: determine utterance situation information corresponding to an utterance situation, and calibrate a frequency weight for a word using the utterance situation information. ### Claim 7 (depends on 6) The artificial intelligence apparatus of claim 6, wherein the utterance situation information includes at least one of a type of the artificial intelligence apparatus, an installation location of the artificial intelligence apparatus, a location of the user, a main user, and a device operation status, and wherein the device operation status includes operation status information of the artificial intelligence apparatus. ### Claim 8 (depends on 7) The artificial intelligence apparatus of claim 7, further comprising a communication circuit configured to communicate with at least one external device, wherein the device operation status further includes operation status information of the at least one external device. ### Claim 9 (depends on 8) The artificial intelligence apparatus of claim 8, wherein the processor is configured to: determine a calibration weight based on the utterance situation information, and calibrate a frequency weight by multiplying a frequency weight by the calibration weight. ### Claim 10 A computer-implemented method of recognizing speech of a user, the method comprising: obtaining, via a microphone, speech data including speech of a user, for each of a plurality of sections of the speech data: calculating a probability for each of a plurality of words, determining, using a speech recognition log, a frequency weight for each of the plurality of words, for each of the plurality of words, calibrating the probability by multiplying the calculated probability by the determined frequency weight, converting the speech data into text by (i) selecting, for each section of the plurality of sections, a word having a highest calibrated probability and (ii) combining the selected words, and performing control based on the text. ### Claim 11 (depends on 10) The computer-implemented method of claim 10, further comprising: normalizing the calibrated probabilities, and converting the speech data into the text by selecting, for each section of the plurality of sections, a word having a highest normalized calibrated probability. ### Claim 12 (depends on 10) The computer-implemented method of claim 10, further comprising: checking a usage frequency of a word using the speech recognition log, and setting, based on usage frequencies of words, a frequency weight of a word to be higher than a frequency weight of another word. ### Claim 13 (depends on 12) The computer-implemented method of claim 12, further comprising: determining a private frequency weight for a word using a private speech recognition log, determining a public frequency weight for a word using a public speech recognition log, and calibrating a probability for a word using the private frequency weight and the public frequency weight, wherein the private speech recognition log includes speech recognition logs in an artificial intelligence apparatus, and wherein the public speech recognition log includes speech recognition logs in an external apparatus. ### Claim 14 (depends on 13) The computer-implemented method of claim 13, further comprising: determining an integrated frequency weight by calculating a weighted sum of a private frequency weight of a word and a public frequency weight of a word, and calibrating a probability using the integrated frequency weight. ### Claim 15 (depends on 10) The computer-implemented method of claim 10, further comprising: determining utterance situation information corresponding to an utterance situation, and calibrating a frequency weight for a word using the utterance situation information. ### Claim 16 (depends on 15) The computer-implemented method of claim 15, wherein the utterance situation information includes at least one of a type of an artificial intelligence apparatus, an installation location of the artificial intelligence apparatus, a location of the user, a main user, and a device operation status, and wherein the device operation status includes operation status information of the artificial intelligence apparatus. ### Claim 17 (depends on 10) The computer-implemented method of claim 10, further comprising: determining a calibration weight based on utterance situation information, and calibrating a frequency weight by multiplying a frequency weight by the calibration weight. ### Claim 18 A non-transitory processor-readable medium having recorded a program for performing a method of recognizing speech of a user, the method comprising: obtaining, via a microphone, speech data including speech of a user, for each of a plurality of sections of the speech data: calculating a probability for each of a plurality of words, determining, using a speech recognition log, a frequency weight for each of the plurality of words, for each of the plurality of words, calibrating the probability by multiplying the calculated probability by the frequency weight, converting the speech data into text by (i) selecting, for each section of the plurality of sections, a word having a highest calibrated probability and (ii) combining the selected words, and performing control based on the text.