25 results about "Speech recognition" patented technology

Methods and apparatus for executing a lighting program to control a plurality of light emitting diodes (LEDs) in response to at least one characteristic of an audio input. In one embodiment, the audio input is digitally processed to determine the at least one characteristic. In other embodiments, control signals for the LEDs are generated in response to a timer and/or input from a user interface, as well as in response to the at least one characteristic of the audio input. In another embodiment, the control signals for the LEDs are generated by a same computer that processes the audio input to transmit signals to speakers to audibly play the audio input. In a further embodiment, a GUI is provided to assist in authoring the lighting program. In another embodiment, the audio signal is processed before being played back. In a further embodiment, the lighting program anticipates changes in the audio input.

The invention provides a method and a system for automatically setting sound effect parameters. The method comprises the steps of firstly, obtaining a music header file of a file currently to be played and determining the corresponding music type according to the music header file; judging whether sound effect parameters need to be changed or not according to the music type, if yes, obtaining the corresponding sound effect parameters of the music type according to a music type and sound effect parameter corresponding table which is set in advance, and decoding and playing the file to be played. By the method provided by the invention, the sound effect parameters are set automatically through analyzing the header file of a music, the sound effect parameters can be changed into the sound effect parameters suitable for the current music type automatically without manual setup of users, and the problem that music quality is not good due to the single sound effect is also avoided.

The invention discloses a conference recording method and device, an intelligent terminal and a storage medium. The method comprises the steps of receiving speech information sent by a participating terminal; acquiring a face image of a speaker corresponding to the participating terminal; matching the face image with reserved information to obtain the first speaker name corresponding to the face image, wherein the reserved information includes a corresponding relationship between conference attendee names and face feature images; and generating conference record on the basis of the first speaker name and the speech information. The name of the speaker is identified through acquiring the face image of the speaker I the participating terminal corresponding to the speech information, the nameand the speech information of the speaker are sorted into a conference record corresponding to the name and speech information of the speaker, thereby intelligently completing the conference record,reducing the labor cost and saving the time cost, and improving the conference recording efficiency.

The invention discloses an anti-tonal modification interference audio fingerprint extraction method. The method comprises the following steps of framing and windowing an audio signal, performing Fourier transform, and collecting Fourier coefficients corresponding to all frames of signals; calculating energy segment indexes corresponding to all the frames of signals by using an energy segment indexcalculation method based on a peak point in the collected Fourier coefficients; and calculating energy segment values by using the energy segment indexes and filtering energy segments by a two-dimensional filtering core to extract fingerprints. By means of the method, the defect that an existing Philips fingerprint cannot resist tonal modification interference can be overcome; and on the premiseof keeping the sub-fingerprint continuity of the original Philips fingerprint and the robustness of various noise interferences, the resistance to the tonal modification interference is improved.

The invention relates to the technical field of artificial intelligence, in particular to the field of word segmentation models, and discloses a word vector definition method and device based on artificial intelligence, equipment and a storage medium, which are used for taking a single word as a minimum structure, endowing each word with a plurality of different meanings in different words, and increasing the accuracy of the meanings of word vectors in sentences. The method comprises the following steps: obtaining a target text, wherein the target text comprises Chinese sentences needing to besubjected to word segmentation; performing word segmentation on the target text to obtain a plurality of words; generating a plurality of corresponding target word vectors according to the pluralityof words; and generating a target word vector according to the plurality of target word vectors and a preset weight strategy, wherein each word in the plurality of words corresponding to the pluralityof target word vectors comprises a word corresponding to the target word vector.

The invention provides a multilayer retrieval method based on word segmentation. The method comprises the steps of obtaining retrieval basic split words of a retrieval text; combining the retrieval basic split words to form retrieval combined split words of the retrieval text; using the retrieval basic split words and the retrieval combined split words of the retrieval text to search for text datain a corpus to obtain a retrieval result of the retrieval text.

An audio signal processing method and a broadcasting output apparatus using the audio signal processing method are provided. The audio signal processing method includes selecting an audio signal to be processed from among a plurality of audio signals in which the same audio data is compressed in different formats, when the plurality of audio signals are received; and processing the selected audio signal. Therefore, it is possible to select a high quality audio signal from among audio signals compressed in various audio formats and to provide a user with the high quality sound.

The invention discloses a domain-specific Chinese event extraction method. The method includes: preprocessing, removing redundant tags, clauses and participles form original data information; dictionary building, namely performing artificial analysis of large-scale news corpus, summarizing domain-specific common characters and building a character dictionary; summarizing common expressions of reporting domain-specific events and building a trigger word based sentence pattern dictionary; sentence pattern matching, namely performing sentence pattern matching on a syntax tree according to the sentence pattern dictionary; event element extraction, namely performing character extraction on the syntax tree according to the character dictionary and performing time and place extraction on the syntax tree according to syntactic features of time and place elements. With the method, the difficulty of quickly acquiring Chinese news information in the big data environment is solved, users can get news events related to keywords by entering the keywords themselves through automatic processing, and great convenience is provided for information acquisition.

The invention discloses an intelligent voice communication processing method and belongs to the voice communication technology field. The method comprises steps that a voice communication signal is acquired; after one of multiple voice processing modules preset in a terminal device is activated, noise reduction of the voice communication signal is carried out through utilizing the voice processingmodule; the multiple voice processing modules have different noise reduction capabilities for the voice communication signal. Before the voice communication signal is acquired, a step of setting themultiple voice processing modules on the terminal device is comprised. The method is advantaged in that the voice communication signal after noise reduction has diversity, and the method can adapt todifferent environments and different users' requirements for voice signals.

The invention discloses a Chinese character learning method and system based on a complex network. According to the method, composition structures of Chinese characters are analyzed, a reasonable algorithm is designed, a composition relation network between the Chinese characters is built and cooperates with the difficulty degree of the Chinese character structures and the application word frequency to calculate the learning cost and benefits of the Chinese characters, and a scientific Chinese character learning sequence is provided. Besides, China is a multi-ethnic country, and the inheritance of minority languages and characters is an important part of modern teaching besides Chinese teaching. Therefore, the method can be popularized to the teaching research process of minority languagesand words. The method can be applied to the teaching process of languages and words and selection, compiling and evaluation of teaching materials, and can also be applied to development of commercialteaching software and APPs so that language learners can learn more scientifically and reasonably, and the maximum learning effect is obtained.

The invention relates to a method and a system for off-line publishing of internet voice frequency content with a literal label. By utilizing the method, the voice frequency content transcribed by the voice frequency module and the word content received by a short message module are combined by adopting a combination manner of transcribing the voice frequency content through dialing and adding the word label through sending short messages, and the word content utilized as the word label is published on an internet server. The system provided by the invention comprises a cellphone, a voice frequency transcription module, a short message module, a combination publishing module and the internet server, wherein, the voice frequency transcription module and the short message module are respectively connected with the cellphone; the combination publishing module is respectively connected with the voice frequency transcription module and the short message module; and the internet server is connected with the combination publishing module. By utilizing the method and system provided by the invention, users are enabled to publish the word information of the voice frequency content in an off-line mode by utilizing the off-line advantages of the communication network, and meanwhile, the word information can be previewed or searched.

The invention discloses a singing scoring system, which comprises a data acquisition module, a controller, a display playing module, a data transmission module, a user terminal and a cloud server; the data acquisition module is used for acquiring user data information and data information in a singing process of a user; the controller is used for processing data information acquired by the data acquisition module to obtain a singing evaluation result, the display and play module is used for displaying the singing evaluation result and the data information acquired by the data acquisition module, and the user terminal and the cloud server communicate with the controller through the data transmission module. The singing evaluation result can be obtained by comprehensively evaluating the singing audio and the singing emotion, and meanwhile, the user can interact with other users through the mobile terminal.

The invention relates to a virtual instrument for real-time musical generation, comprising a musical rule set unit (7) for defining musical rules, a time constrained pitch generator (9) for synchronizing generated music, an audio generator (11) for generating audio signals (10), wherein the rule definitions describe real-time morphable music parameters, and said morphable music parameters are controllable directly by the real-time control signal (2). With this virtual instrument, the user can create new musical content in a simple and interactive way regardless of the level of musical training obtained before using said instrument.

The embodiment of the invention provides a method and device for splitting phrases, and relates to the technical field of information processing. The method comprises the following steps: acquiring a phrase to be split; extracting words in the phrase to be split according to the default strategy; splitting the words from the phrase to be split when the words and the words in the preset vocabulary database are consistent. According to the phrase splitting method and device provided by the invention, the words in the phrase to be split are extracted according to the default strategy, and the words are split from the phrase to be split when the words and the words in the preset vocabulary database are consistent, so that the phrase-splitting error caused by the word determination error is avoided, thus the phrase-splitting accuracy is improved.

The invention provides a method and a device for classifying sentences based on TF-IDF (term frequency-inverse document frequency). The method comprises the following steps: obtaining news wartime sentences, and preprocessing the sentences; extracting TF-IDF (term frequency-inverse document frequency) features of the preprocessed sentences; and classifying the news combat sentences based on the TF-IDF features of the sentences and a preset theme category. In this way, the sentences are preprocessed, and the sentences are classified according to the TF-IDF features and the preset theme categories, so that the classification accuracy is improved.

A method, an electronic device, and a recording medium for compensating an in-ear audio signal are provided. The method is applicable to the electronic device having a processor. In the method, an in-ear audio signal transmitted through an inner ear when a user speaks is captured by using an in-ear microphone, and an outer audio signal transmitted through air when the user speaks is captured by using an outer microphone in a training stage. Then, a machine learning model of audio signals is established for an objective function and is trained by using the in-ear audio signal and the outer audio signal. Finally, the in-ear audio signal captured by the in-ear microphone is converted into a compensated audio signal by using the trained machine learning model in an online stage, and the compensated audio signal is output.

The embodiment of the invention discloses an audio noise reduction method and device, electronic equipment and a storage medium. The method comprises the following steps: acquiring an audio to be denoised; obtaining voiceprint features of a user corresponding to the to-be-denoised audio, wherein the voiceprint features are obtained after feature extraction is performed on registered audio of the user; and inputting the to-be-denoised audio and the corresponding voiceprint features into a denoising model to obtain a denoised audio corresponding to the user, the denoising model being used for obtaining the denoised audio corresponding to the user from the to-be-denoised audio according to the voiceprint features. And the noise reduction model finds the audio corresponding to the voiceprint feature from the audio to be subjected to noise reduction in combination with the voiceprint feature of the user, and outputs the audio corresponding to the voiceprint feature, thereby completing noise reduction of the audio to be subjected to noise reduction. The noise-reduced audio of the user corresponding to the voiceprint feature and not containing interference audio of other users and background noise is obtained, and the noise reduction effect on the audio to be subjected to noise reduction of the user is improved.

The invention discloses an online translation model selection method for statistical machine translation, which includes two stages of training and translation. The training stage includes: collecting bilingual parallel corpora, and dividing the bilingual parallel corpus into different sub-corpora according to types; The corpus trains the candidate translation model; builds an index for the sub-corpus, and obtains the corpus index file; the translation stage includes: inputting the text to be translated, and retrieving sentences similar to the sentences in the text to be translated from the corpus index file; The candidate translation model corresponding to the sub-corpus where the sentence is located, selects the final translation model from all the candidate translation models; translates the input text to be translated according to the final translation model, and obtains the final translation result. The invention well solves the problem that the statistical machine translation system cannot adapt to input texts in different fields, and can effectively improve the translation quality of the statistical machine translation system.

The disclosed method comprises: a. listing the content to be record; b. selecting the beta-wave music with beat as 70-200/min; c. with the beat of beta-music, recording the contents both of the language and opposite Chinese on the audiovisual product with the first character of content reading out along the beta-wave music and other characters reading out continually. This invention is benefit to language learning.