223results about How to "Improve speech recognition accuracy" patented technology

A speech recognition system having multiple recognition vocabularies, and a method of selecting an optimal working vocabulary used by the system are disclosed. Each vocabulary is particulary suited for recognizing speech in a particular language, or with a particular accent or dialect. The system prompts a speaker for an initial spoken response; receives the initial spoken response; compares the response to each of a set of possible responses in an initial speech recognition vocabulary to determine a response best matched in the initial vocabulary. A working speech recognition vocabulary is selected from a plurality of speech recognition vocabularies, based on the best matched response.

The invention discloses a voice interaction method and a voice interaction device. The method comprises steps: after a voice recognition text is received, the voice recognition text is distributed to each service respectively, and semantic understanding is carried out respectively; and based on the obtained semantic understanding result and the application state of a client, confidence ranking is carried out, a semantic understanding result with the highest confidence is acquired, and a response is given to the semantic understanding result with the highest confidence. As confidence ranking of semantic understanding results on multidimensional-based information not only considers the matching degree between a semantic understanding result and each service but also considers the application state of the client, for example, whether the client is in a navigation state or a music listening state, the application of the client and the application state are possibly objects to be processed by voice interaction, semantic understanding on the multidimensional-based information can effectively enhance the accuracy of judging the service belonging, the accuracy of man-machine interaction semantic understanding is improved, and the user experience is enhanced.

A speech recognition system comprising: a language identification unit for identifying the language of a text item entry; at least one separate pronunciation modelling unit including a phoneme set and pronunciation model for at least one language; means for activating the pronunciation modelling unit including the phoneme set and pronunciation model for the language corresponding to the language identified in the language identification unit for obtaining a phoneme transcription for the entry; and a multilingual acoustic modelling unit for creating a recognition model for the entry.

Within an interactive voice response system, a method of automatically disambiguating results presented to a user can include determining the identity of a user within an interactive voice response session, receiving user inputs specifying selections in an interactive voice response menu hierarchy, and storing historical information specifying the user selections within a profile associated with the identity of the user. For at least one subsequent input from the user, identifying the historical information associated with the identity of the user and using the historical information to reduce a number of possible selections in the interactive voice response menu hierarchy which are presented to the user.

The invention discloses a speech recognition method used for recognizing speech of users and generating speech recognition results. The method includes following steps: a speech obtaining step, obtaining the speech of the user; a speech recognition step, recognizing the obtained speech as text data and regarding the data as an initial speech recognition result; and an error correction step, checking an error correction list which records the relevance of a plurality of original vocabularies-deviation vocabularies, performing error correction processing if a deviation vocabulary in the relevance of a certain original vocabulary-deviation vocabulary in the error correction list existing in the initial speech recognition result, and replacing the vocabulary which is the same with the deviation vocabulary in the initial speech recognition result into the original vocabulary in the relevance of the original vocabulary-deviation vocabulary. In this way, the speech recognition result after error correction is obtained and regarded as the speech recognition result generated by the speech recognition method. The invention also discloses a speech assessment method based on the speech recognition method, and a corresponding speech recognition system and a speech assessment system.

A speech recognition system and method are provided to correctly distinguish among multiple interpretations of an utterance. This system is particularly useful when the set of possible interpretations is large, changes dynamically, and/or contains items that are not phonetically distinctive. The speech recognition system extends the capabilities of mobile wireless communication devices that are voice operated after their initial activation.

A distributed voice recognition system (500) and method employs principles of bottom-up (i.e., raw input) and top-down (i.e., prediction based on past experience) processing to perform client-side and server-side processing by (i) at the client-side, replacing application data by a phonotactic table (504); (ii) at the server-side, tracking separate confidence scores for matches against an acoustic model and comparison to a grammar; and (iii) at the server-side using a contention resolver (514) to weight the client-side and server-side results to establish a single output which represents the collaboration between client-side processing and server-side processing.

[Object] An object is to provide an easy-to-use speech processing system attaining higher accuracy of speech recognition.

[Solution] Receiving a speech utterance, the speech processing system performs speech recognition and displays a text 158 of the recognition result. Further, the speech processing system translates the recognition result in accordance with settings to a text 176 of another language and displays and synthesizes speech of the translated result. Further, the speech processing system selects utterance candidates having high possibility to be uttered as the next utterance and having high translation and speech recognitions scores, using outputs of various sensors at the time of utterance, a pre-trained utterance sequence model and translation and speech recognition scores of utterance candidates, and recommends utterance candidates in the form of an utterance candidate recommendation list 192. A user can think of what to say next using the utterances in utterance candidate recommendation list 192 as hints.

The invention provides a speech recognition method and a speech recognition device based on a Chinese and English mixed dictionary. The speech recognition method comprises the steps of: acquiring the Chinese and English mixed dictionary marked by the International Phonetic Alphabet IPA, wherein the Chinese and English mixed dictionary comprises a Chinese dictionary and an English dictionary corrected by means of the Chinese dictionary; training the model by regarding the Chinese and English mixed dictionary as a training dictionary, regarding a layer of convolutional neural network CNN plus five layers of Long Short-Term Memory (LSTM) network as a model, regarding status of the IPA as a target and regarding a connectionist time classifier CTC as a training criterion, so as to obtain a trained CTC acoustic model; and combining with the trained CTC acoustic model for performing speech recognition on a Chinese and English mixed language. According to the speech recognition method and the speech recognition device, the Chinese and English mixed dictionary comprising the Chinese dictionary and the English dictionary corrected by means of the Chinese dictionary is adopted for training, the English word coverage is comprehensive and Chinglish can be recognized, and the accuracy degree of Chinese and English mixed language recognition is further improved by combining the application of the CTC acoustic model.

The invention discloses a speech recognition method and a speech recognition device. A specific embodiment of the method comprises the following steps: segmenting speech information to be recognized into multiple frames of speech segments; performing acoustic model scoring and language model score-checking on the speech segments frame by frame through a preset decoding network; and determining a word sequence corresponding to at least one decoding path in the decoding network as the speech recognition result based on the score result, wherein first language model score-checking and second language model score-checking are carried out in sequence during language model score-checking on a frame of speech segment. By implementing the technical scheme, accurate and efficient speech recognition is realized.

A unified speech input dialogue interface, and a distributed multiple application-dependent language processing unit system with the unified speech recognition function and the unified dialogue interface are provided. The system not only provides a convenient user's environment, but also enhances the whole performance of speech recognition. The distributed multiple application-dependent language processing unit system uses a speech input interface so that the user can be familiar with a simple, unified interface. The system also improves the speech recognition accuracy and enhances the convenience of use by self-learning personalized dialogue model.

Methods and systems are provided for improving the accuracy of a speech recognition system in recognizing alphabetic character input. If spoken alphabetic characters are erroneously recognized by a speech recognition system, a user may reenter the alphabetic characters using DTMF key tones on the user's telephone keypad to assist the speech recognition system in determining the correct input. If the input is originally input using DTMF key tones, the user may reenter the input as spoken alphabetic characters to assist the system to identify the correct input from the combinations of input that may be associated with the DTMF key tone entry.

The invention belongs to the technical field of signal processing in the electronic industry, and aims at solving a problem that the discrimination performance of an acoustic model of a target language with just a little mark data is low. In order to solve the above problem, the invention provides a small data speech acoustic modeling method in speech recognition, and the method comprises the steps: carrying out the adversarial training of the acoustic features of many languages through a language adversarial discriminator, so as to build a multi-language adversarial bottleneck network model;taking the acoustic features of a target language as the input of the multi-language adversarial bottleneck network model, so as to extract the bottleneck features which is irrelevant to the language;carrying out the fusion of the bottleneck features which is irrelevant to the language with the acoustic features of the target language, so as to obtain fusion features; carrying out the training through the fusion features, so as to build an acoustic model of the target language. The method effectively irons out the defects, caused by a condition that the bottleneck information comprises the information correlated with the language, of the unremarkable improvement of the recognition performance of the target language and even the negative migration phenomenon in the prior art, thereby improving the voice recognition precision of the target language.

The invention relates to the technical field of voice recognition, provides a voice recognition method and device and electronic equipment, which aim to solve the problem of relatively low voice recognition accuracy. The method comprises the following steps of acquiring a to-be-recognized voice, performing feature extraction on the to-be-recognized voice to obtain voice feature information, determining a target character sequence corresponding to the voice feature information according to the target acoustic model and the target language model, wherein the target language model comprises a first language model and a second language model, the first language model is obtained by performing language model training through a command word training text of a first scene, and the second languagemodel is obtained by performing language model training through a first text training set. In the voice recognition process, two language models are adopted, and the first language model is obtainedby performing language model training through the command word training text of the first scene, so that the recognition capability of the first language model for related command words in the first scene can be enhanced, and the voice recognition accuracy can be improved.

Inputting speech includes receiving feature information obtained by a client, the feature information comprising speech signals and user feature image signals, recognizing first candidate recognition data matching the user feature image signals, determining target recognition data based at least on the first candidate recognition data, and outputting the target recognition data.

The invention relates to a scheme of realizing network speech recognition in an English oral language machine examination system. According to the scheme, traditional spectral subtraction (SS) noise reduction technology and cepstral mean normalization (CMN) noise reduction technology are improved, combined with a probability scale DP identification method of a continuous state hidden Markov model(HMM), the invention provides a network speech recognition scheme of unspecified people in an English network examination system, and by utilizing the scheme, a network speech recognition apparatus in a physical environment is realized. By employing the above method, an SS method with input amplitude spectrum self-adapting and a CMN method based on progressive adaptive mode MAP algorithm are combined, and influence of ambient noise on an identification system is substantially reduced. Simultaneously, according to the scheme, based on a traditional DP method, by utilizing a DP algorithm of probability scale, recognition is carried out, thus a DSP speech recognition apparatus can be applied to speech recognition of unspecified people of different outdoor occasions, and a recognition system scope and recognition precision are raised.

A speech processing device includes a speech recognition unit configured to sequentially recognize recognition segments from an input speech, a reverberation influence storage unit configured to store a degree of reverberation influence indicating an influence of a reverberation based on a preceding speech to a subsequent speech subsequent to the preceding speech and a recognition segment group including a plurality of recognition segments in correlation with each other, a reverberation influence selection unit configured to select the degree of reverberation influence corresponding to the recognition segment group which includes the plurality of recognition segments recognized by the speech recognition unit from the reverberation influence storage unit, and a reverberation reduction unit configured to remove a reverberation component weighted with the degree of reverberation influence from the speech from which at least a part of recognition segments of the recognition segment group is recognized.

The invention discloses a complex field speech enhancement method and system based on a generative adversarial network and a medium. The complex field speech enhancement method comprises the implementation steps that a noisy speech is acquired; the speech is expressed with Cartesian coordinates after being subjected to Fourier transformation to obtain a noisy real number spectrum and imaginary number spectrum; the noisy real number spectrum and imaginary number spectrum are input into a generator, which completes training in advance, of the generative adversarial network, and a denoised real number spectrum and imaginary number spectrum of a pure speech are obtained; and the real number spectrum and imaginary number spectrum of the pure speech are made into a clean speech based on inverseFourier transformation. Through the method, noise can be better removed from a speech signal to generate the clean speech, the problem that it is difficult to predict a phase is effectively solved, the auditory effect of the enhanced speech can be effectively improved, and the speech recognition accuracy rate of a speech recognition system in a noise environment can be effectively increased.