30results about How to "Accurate Speech Recognition Results" patented technology

The invention proposes a refrigerator, a voice interaction system, a voice interaction method, a computer device and a computer readable storage medium, wherein the voice interaction system comprises a voice acquiring module used to acquire and pre-process the voice signals and to transmit the pre-processed voice signals to a voice recognition module; the voice recognition module used to convert the processed voice signals into text information and to transmit the text information to a processing module and a semantics understanding module respectively; the semantic understanding module used to convert the text information into semantic result and to transmit the semantic result to the processing module; and the processing module used to check and combine related information base according to the text information and the semantic result to execute voice interaction tasks corresponding to the voice signals. According to the invention, interactive scenes are provided in advance so as to understand the user's languages more accurately, to realize more natural human-machine interactions, to increase the interaction experience and to make the refrigerator more intelligent.

The invention discloses a speech-recognition text classification method and device. The method comprises the steps of respectively collecting training texts and training speeches identical with contents of the training texts for all service classes according to service class types, decoding the training speeches to obtain a work confusion network of the training speeches, extracting text characteristics of the training texts according to the training texts and the work confusion network, training a support vector machine classifier in a set according to the text characteristics, and using the trained support vector machine classifier to classify the texts. The speech-recognition text classification method and device converts a word graph network into the word confusion network suitable for text classification. After confusion words contained in the word confusion network are converted into text characteristics, a support vector machine algorithm is utilized to carry out text classification based on the confusion words. Thus, more accurate classification results can be obtained, and the accuracy of speech-recognition text classification is improved.

The embodiment of the invention provides a speech recognition method, a speech recognition device and electronic equipment, and belongs to the technical field of computers. The speech recognition method comprises the steps of obtaining the feature information of a current to-be-recognized speech signal; processing the feature information through an encoder to obtain M speech feature vectors, M being a positive integer greater than or equal to 1; processing the M speech feature vectors through a CTC model, obtaining N pronunciation units in the M speech feature vectors, wherein N is a positiveinteger smaller than or equal to M and larger than or equal to 1; processing the M speech feature vectors and the N pronunciation units through an attention model, generating Y context feature vectors, wherein Y is a positive integer larger than or equal to N; and processing the Y context feature vectors through a decoder network to obtain a recognition result of the current to-be-recognized speech signal. According to the technical scheme of the embodiment of the invention, the accuracy of speech recognition can be improved.

The invention provides a speech control method and a speech control device, a balance car control method and a balance car control device, and a balance car, wherein the speech control method is executed by a controller. The method comprises the following steps: receiving speech signals transmitted from a user by virtue of a speech acquisition device; in accordance with a preset rubbish word list, filtering the speech signals, wherein words, which are irrelevant to target identification key words, are saved in the rubbish word list; comparing the filtered speech signals with the target identification key words which are saved in a preset key word list, so that a speech recognition result is generated, wherein the speech recognition result contains control command; researching command codes, which correspond to the control commands, from a preset command code list; and transmitting the command codes to an executor, so that the executor can control the balance car to move in accordance with the speech recognition result. According to the invention, the speech recognition result can be rapidly and accurately acquired, so that the balance car can be controlled in accordance with the result; therefore, efficiency and precision of controlling the balance car can be improved.

The embodiment of the invention provides a voice recognition method and device. The method includes the steps that a first acoustic model is obtained, the first acoustic model represents the mapping relation of input voice signals and a preset state, and the preset state is a basic element of text corresponding to the voice signals; the first voice signal is received; the probability value of the preset state is determined according to the first voice signal and a second acoustic model; the second acoustic model is determined according to a second input voice signal and the first acoustic model, and the first voice signal and the second voice signal are input by the same user; the voice signals include the first voice signal and the second voice signal; the probability value of the preset state corresponding to the first voice signal is decoded as first text according to a preset voice model and a dictionary. By means of the voice recognition method and device, recognition precision of voice input by specific users can be improved.

The invention provides a speech recognition method and device, and the method comprises the steps: carrying out the clustering processing of a training data feature vector, and obtaining a cluster training data feature vector; carrying out the interpolation calculation of a to-be-recognized data feature vector through the cluster training data feature vector; inputting the to-be-recognized data feature vector into a speech recognition model after the interpolation calculation, so as to achieve the adaptive adjustment of the speech recognition model. The method can improve the self-adaption accuracy and processing efficiency of speech recognition.

The invention provides a subtitle generation method and device, equipment and a storage medium. The method comprises the following steps: when receiving a subtitle generation triggering operation forat least one audio track in a target audio and video file, respectively carrying out voice recognition on audio data on each audio track in the at least one audio track to obtain a text fragment corresponding to each audio track; and then, based on the text fragment corresponding to each audio track, generating subtitles of the target audio and video file. Compared with the whole voice recognitionfor the audio data on all the audio tracks, the recognition precision is improved; according to the embodiment of the invention, independent voice recognition is carried out on the audio data on eachaudio track, the influence between the audio tracks is avoided, a more accurate voice recognition result can be obtained, and the accuracy of subtitles generated based on the voice recognition resultis further improved.

The invention discloses a voice recognition method, device and system, and a computer readable storage medium. The method comprises the steps: carrying out the voice recognition of to-be-recognized voice data, and obtaining each first candidate result of the to-be-recognized voice data; obtaining feature data corresponding to each first candidate result, and inputting each piece of feature data into a pre-trained sorting model to obtain the sorting result of each first candidate result; and selecting a target candidate result from the first candidate results as the voice recognition result ofthe to-be-recognized voice data according to the sorting result. Compared with a linear score combination mode set according to human experience, a more accurate sorting result can be obtained, so that a more accurate voice recognition result can be obtained.

The invention discloses a translation method and device, electronic equipment and a readable storage medium, and relates to a translation technology. According to the embodiment of the invention, according to associated information of a to-be-translated content, at least one knowledge content is obtained, wherein each knowledge content in the at least one knowledge content comprises contents of afirst language type and contents of a second language type. According to the method and the device, the translation result of the to-be-translated content can be obtained by utilizing the at least oneknowledge content, and the at least one knowledge content obtained in advance is used as the global information of the translation task, so that the consistency of the translation results of the sameto-be-translated content can be ensured, and the quality of the translation result is improved.

The invention provides a speech recognition method based on video analysis, and the method comprises the steps: carrying out the cutting and graying of each video frame in target video data, and obtaining an input image corresponding to each video frame; performing face detection processing on all the input images to obtain a face detection frame; screening the face detection frames corresponding to all the input images, and taking the face detection frame corresponding to each video frame conforming to a preset face screening rule as a final face detection frame of the frame; and calculating a feature result corresponding to each final face detection frame according to the lip contour and the face key points in the final face detection frames, inputting a plurality of feature results into a preset speaking recognition model for recognition, and obtaining a speaking recognition result corresponding to the to-be-recognized person. According to the speech recognition method based on video analysis, the obtained speech recognition result is more accurate, and the speech recognition method can adapt to different forms when the person to be recognized speaks.

The invention discloses a voice processing method and device based on household appliances. The method comprises the following steps: the household appliance receiving voice information, and dividingthe voice information into at least the following two parts in a voice recognition process of the voice information: first voice information corresponding to a reasonable statement and second voice information corresponding to an unreasonable statement; obtaining a first recognition result corresponding to the first voice information, and converting the second voice information into a reasonable statement; performing voice recognition on the converted second voice information to obtain a second recognition result; and generating an instruction statement for controlling the household appliancebased on the first recognition result and the second recognition result. The technical problem that in the prior art, household appliance equipment generates statements according to the natural language, and the accuracy is low is solved.

The invention relates to an audio data collecting method and equipment used for reducing audio data which is missed during collection and increasing data integrity. The method includes: when a first trigger, a second trigger and a third trigger are used to detect audio signals and the fact that the energy of the audio signal detected by the first trigger is larger than or equal to the triggering threshold of the first trigger is determined, determining that the first trigger is triggered, wherein the triggering threshold of the first trigger is smaller than a first preset threshold; enabling the data channel of a storage so as to use the storage to collect the audio data input into an audio input system.

The invention discloses a speech recognition method and device in a fast application, equipment and a storage medium, which are applied to the field of computers. The method comprises the following steps of acquiring a first speech signal, wherein the first speech signal is a speech signal received by a first fast application in a terminal in a running state; sending the first speech signal to a first speech recognition server, wherein the first speech recognition server is a speech recognition server corresponding to the first fast application, and the speech recognition servers correspondingto the at least two fast applications are different; and receiving the speech recognition result sent by the first speech recognition server. According to the method, the speech signal acquired by the terminal is sent to the speech recognition server specified by the currently running fast application, so that the speech recognition result is more accurate.

The invention discloses a speech recognition method and device, electronic equipment and a storage medium. The speech recognition method comprises the steps of: carrying out language classification processing on speech information, and obtaining language type information; analyzing and processing the statement relationship of the speech information to obtain language information for describing the statement relationship in the speech information; performing speech feature extraction processing on the speech information, in the process of obtaining speech features, inputting the speech information into a speech coding model, carrying out acoustic feature extraction processing on the speech information in the speech coding model, adopting obtained acoustic features as the speech features; and performing speech recognition processing according to the language type information, the language information and the speech features to obtain a speech recognition result. According to the speech recognition method and device, the electronic equipment and the storage medium of the invention, the accuracy of speech recognition can be improved.

The invention relates to the technical field of voice recognition, in particular to a voice recognition method and device based on waveform simulation, and the method comprises the following steps: 1, collecting a to-be-recognized voice signal, and carrying out the waveform envelope simulation of the to-be-recognized voice signal, so as to obtain the envelope simulation waveform of the to-be-recognized voice signal; and 2, carrying out waveform simplification simulation on the envelope simulation waveform to obtain a simplified simulation waveform of the envelope simulation waveform . According to the method, waveform envelope simulation is carried out on a voice signal to obtain a simplified waveform of the voice signal, the simplified waveform is further simplified to obtain the most simplified waveform which is easier to recognize, and the simplified waveform is subjected to voice recognition through image matching recognition, so that the method is different from existing voice recognition, the recognition accuracy is higher, and meanwhile, the simplified waveform is used for recognition, so that the efficiency is improved and is equivalent to that of a conventional voice recognition method.

The invention relates to an audio data collecting method and equipment used for reducing audio data which is missed during collection and increasing data integrity. The method includes: when a first trigger, a second trigger and a third trigger are used to detect audio signals and the fact that the energy of the audio signal detected by the first trigger is larger than or equal to the triggering threshold of the first trigger is determined, determining that the first trigger is triggered, wherein the triggering threshold of the first trigger is smaller than a first preset threshold; enabling the data channel of a storage so as to use the storage to collect the audio data input into an audio input system.

The invention discloses a speech recognition adjustment method and device. The method includes: step 101, receiving a first speech from a user; step 102, determining at least one candidate recognition result based on the first speech, and determining at least one candidate recognition result based on the first recognition The strategy selects a candidate recognition result from the at least one candidate recognition result to form a first speech recognition result; step 103, receiving a second speech from the user; step 104, responding to the second speech, based on the second recognition A strategy selects a candidate recognition result from other candidate recognition results in the at least one candidate recognition result to form a second speech recognition result; wherein the second recognition strategy is different from the first recognition strategy. Through the method of the present invention, more accurate speech recognition results can be provided, the correction process of speech recognition is accelerated, the error rate of speech recognition is reduced, and user experience is improved.

The invention discloses a voice processing method and device based on household appliances. The method comprises the following steps: the household appliance receiving voice information, and dividingthe voice information into at least the following two parts in a voice recognition process of the voice information: first voice information corresponding to a reasonable statement and second voice information corresponding to an unreasonable statement; obtaining a first recognition result corresponding to the first voice information, and converting the second voice information into a reasonable statement; performing voice recognition on the converted second voice information to obtain a second recognition result; and generating an instruction statement for controlling the household appliancebased on the first recognition result and the second recognition result. The technical problem that in the prior art, household appliance equipment generates statements according to the natural language, and the accuracy is low is solved.

Embodiments of the present invention provide a voice recognition method, device, equipment, and storage medium. Wherein, the voice recognition method may include: acquiring multiple wake-up voice signals from multiple locations; performing sound source localization on the multiple wake-up voice signals to determine the position of the wake-up voice; The audio signal of the position is suppressed to obtain the signal to be recognized; voice recognition is performed on the signal to be recognized. By first determining the wake-up voice position, audio signals at other positions can be suppressed, thereby maintaining the validity of the voice at the wake-up voice position, reducing the impact of noise signals at other positions on speech recognition, and reducing the interference to the wake-up voice position.

The invention provides a voice recognition method and device, electronic equipment and a storage medium, and belongs to the technical field of voice recognition. The method comprises the steps: acquiring an audio signal and a vibration signal, the vibration signal is a signal generated by skin vibration of a sound production part of a first object, wherein the time interval between the acquisition time of the audio signal and the acquisition time of the vibration signal is within a preset range; respectively determining an association feature between the audio signal and the vibration signal, an audio feature of the audio signal and a vibration feature of the vibration signal; fusing the associated features, the audio features and the vibration features to obtain fused features; and performing voice recognition on the fused features. According to the voice recognition method and device, voice recognition is carried out in combination with the vibration signals and the correlation characteristics between the audio signals and the vibration signals, even in a noisy environment, interference of noise on voice recognition can be reduced, and the accuracy of voice recognition is improved.

The present invention provides a refrigerator, a voice interaction system, a voice interaction method, a computer device, and a computer-readable storage medium, wherein the voice interaction system includes: a voice collection module for collecting and preprocessing voice signals, and preprocessing The final speech signal is sent to the speech recognition module; the speech recognition module is used to convert the preprocessed speech signal into text information, and sends the text information to the processing module and the semantic understanding module respectively; the semantic understanding module is used to convert the text The information is converted into a semantic result, and the semantic result is sent to the processing module; the processing module is used to query and combine the associated information base to perform the voice interaction task corresponding to the voice signal according to the text information and the semantic result. The present invention pre-sets the interaction scene, understands the user's language more accurately, realizes more natural human-computer interaction, improves the interaction experience, and improves the intelligence of the refrigerator.

The invention relates to an audio data collecting method and equipment used for reducing audio data which is missed during collection and increasing data integrity. The method includes: when a first trigger and a second trigger are used to detect audio signals and the fact that the energy of the audio signal detected by the first trigger is larger than or equal to the triggering threshold of the first trigger is determined, determining that the first trigger is triggered, wherein the triggering threshold of the first trigger is smaller than a first preset threshold; enabling the data channel of a storage so as to use the storage to collect the audio data input into an audio input system.

The invention relates to an equipment nameplate dual intelligent identification method, a storage medium and a terminal. The method comprises the following steps: S1, obtaining an equipment nameplate image; s2, judging whether the definition of the equipment nameplate image can be subjected to character recognition or not, if yes, recognizing characters in the equipment nameplate image to obtain equipment nameplate content, ending, and if not, executing the step S3; s3, acquiring equipment nameplate voice data; and S4, performing voice recognition on the equipment nameplate voice data to obtain equipment nameplate content, and ending. Compared with the prior art, rapid and accurate identification of the nameplate of the power equipment in a complex power environment is realized.

The invention discloses a text classification method and device for voice recognition, wherein the method includes: collecting training text and training voice consistent with the content of the training text for each business category according to the type of the business category; Decode to obtain the word confusion network of the training speech; according to the training text and the word confusion network, extract the text feature of the training text; according to the text feature, support vector machine classifier in the training set; by the trained support vector machine classifier Classify text. The text classification method and device for speech recognition of the present invention convert the word graph network into a word confusion network suitable for text classification, convert the confused words contained in the word confusion network into text features, and use the support vector machine algorithm to perform a method based on The text classification of confusing words can get more accurate classification results and improve the accuracy of speech recognition text classification.

The invention relates to a pronunciation error detection method and device, computer equipment and a storage medium. The method comprises the following steps: acquiring voice information and standard pronunciation text information corresponding to the voice information; performing voice recognition according to the voice information and the standard pronunciation text information to obtain predicted pronunciation text information corresponding to the voice information; and detecting pronunciation error of the voice information according to the predicted pronunciation text information and the standard pronunciation text information. By adopting the method, the pronunciation error detection accuracy can be improved.

The application discloses a translation method, device, electronic equipment and readable storage medium, and relates to translation technology. In this embodiment of the present application, at least one knowledge content is obtained according to the associated information of the content to be translated, and each knowledge content in the at least one knowledge content includes the content of the first language type and the content of the second language type, so that the at least one knowledge content can be used One piece of knowledge content, to obtain the translation result of the content to be translated, because at least one knowledge content obtained in advance is used as the global information of this translation task, so that the same content to be translated can be guaranteed, and the translation results are consistent, thereby improving translation. the quality of the results.

The invention relates to an audio data collecting method and equipment used for reducing audio data which is missed during collection and increasing data integrity. The method includes: when a first trigger and a second trigger are used to detect audio signals and the fact that the energy of the audio signal detected by the first trigger is larger than or equal to the triggering threshold of the first trigger is determined, determining that the first trigger is triggered, wherein the triggering threshold of the first trigger is smaller than a first preset threshold; enabling the data channel of a storage so as to use the storage to collect the audio data input into an audio input system.

This application discloses a voice recognition method, device, electronic equipment and storage medium. The specific implementation plan is: perform language classification processing on the voice information to obtain language information; analyze and process the sentence relationship on the voice information to obtain the The linguistic information of the sentence relationship in the speech information; the speech feature extraction process is carried out on the speech information, and in the process of obtaining the speech features, the speech information is input into the speech coding model, and the speech is processed in the speech coding model The information is subjected to the extraction process of the acoustic feature, and the obtained acoustic feature is used as the speech feature; the speech recognition process is performed according to the language type information, the language information and the speech feature, and a speech recognition result is obtained. By adopting the application, the accuracy rate of speech recognition can be improved.

The embodiment of the invention discloses a sorting method and device for voice recognition candidate results. The method comprises the steps: acquiring candidate result information of voice recognition; combining the identified candidate result information in pairs to generate at least one candidate combination; identifying common erroneous words of the candidate combinations; replacing a commonerror word included in each candidate result information in the candidate combination with a marker word, replacing word information of the common error word with marker word information, obtaining areplacement candidate result corresponding to each candidate result information, and obtaining a to-be-classified combination; utilizing a pre-trained binary classification model to obtain the advantages and disadvantages of two replacement candidate results in each to-be-classified combination; and sorting the candidate result information according to the advantages and disadvantages of the two replacement candidate results in each to-be-classified combination. By obtaining the advantages and disadvantages of every two replacement candidate results, the advantages and disadvantages sequence of the replacement candidate result information can be determined, and an accurate voice recognition result is obtained.