53 results about "Speech retrieval" patented technology

A system for indexing and searching speech content, the system includes two distinct stages, a speech indexing stage (100) and a speech retrieval stage (200). A phone lattice (103) is generated by passing speech content (101) through a speech recogniser (102). The resulting phone lattice is then processed to produce a set of observed sequences Q=(Θ,i) where Θ are the set of observed phone sequences for each node i in the phone lattice. During the retrieval stage (200), a user first inputs a target word (205) into the system, which is then reduced to a target phone sequence P=(p₁, p₂, . . . , p_N) (207). The system then compares target sequence P with the set of observed sequences Q (208), suitably by scoring each observed sequence against the target sequence using a Minimum Edit Distance (MED) calculation to produce a set of matching sequences R (209).

The invention provides a method and a device for establishing a language model and a method and a device for recognizing voice. The method for establishing the language model comprises the following steps of: acquiring timeliness search corpora; carrying out language model training by utilizing the acquired timeliness search corpora to obtain a timeliness language model; and fusing the timeliness language model with a background language model to obtain the final recognition language model, wherein the background language model is used for describing a long-term retrieval behaviour of a user. Through adopting the recognition language model obtained by the method provided by the invention, when the user sends a voice retrieval request for an emergency, the request of the user can be accurately recognized, so that a reliable retrieval result can be provided for the user.

The invention provides an information retrieval system and a method based on bilingual phonetic query. The system comprises: a character extractor which is used for converting a phonetic signal from PCM waveform into an MFCC characteristic parameter and outputting a compressed MFCC data flow after the noise suppression and frame compressing processing; a bilingual phoneme recognizer which is used for receiving the compressed MFCC data flow and executing the speech recognition by automatically converting English or Chinese speech into text phoneme series; a bilingual character phoneme converter which is used for converting character level series of available contents used for the speech retrieval of MP3 files in an MP3 ID3 label database into phoneme level series used as referential phoneme series; a vocabulary comparator which is used for comparing the recognizing phoneme series generated from the bilingual phoneme recognizer with the referential phoneme series generated from the bilingual character phoneme converter and outputting the most related front N referential phoneme series.

Disclosed are a speech retrieval apparatus and a speech retrieval method for searching, in a speech database, for an audio file matching an input search term by using an acoustic model serialization code, a phonemic code, a sub-word unit, and a speech recognition result of speech. The speech retrieval apparatus comprises a first conversion device, a first division device, a first speech retrieval unit creation device, a second conversion device, a second division device, a second speech retrieval unit creation device, and a matching device. The speech retrieval method comprises a first conversion step, a first division step, a first speech retrieval unit creation step, a second conversion step, a second division step, a second speech retrieval unit creation step, and a matching step.

Methods of document expansion for a speech retrieval document by a recognizer. A database of vectors of automatic transcriptions of documents is accessed and the vectors are truncated by removing all terms that are not recognizable by the recognizer to create truncated vectors. Terms in the vectors are then weighted to associate the truncated vectors with the untruncated vectors. Terms not recognized by the recognizer are then added back to the weighted, truncated vectors. The retrieval effectiveness may then be measured.

The invention relates to the technical field of speech retrieval, and discloses a speech retrieval method and a speech retrieval system. The method comprises the following steps that retrieval keywords input by users are received; the retrieval keywords are subjected to single word segmentation to obtain single word segmentation phases; the retrieval keywords are expanded according to the single word segmentation phases to generate a key phase picture structure; and phases on each arch in the key phase picture structure are retrieved according to a pre-built index database to obtain retrieval results. The method and the system provided by the invention are utilized, and the effectiveness and the comprehensiveness of the retrieval results can be improved.

A method for speech retrieval includes acquiring a keyword designated by a character string, and a phoneme string or a syllable string, detecting one or more coinciding segments by comparing a character string that is a recognition result of word speech recognition with words as recognition units performed for speech data to be retrieved and the character string of the keyword, calculating an evaluation value of each of the one or more segments by using the phoneme string or the syllable string of the keyword to evaluate a phoneme string or a syllable string that is recognized in each of the detected one or more segments and that is a recognition result of phoneme speech recognition with phonemes or syllables as recognition units performed for the speech data, and outputting a segment in which the calculated evaluation value exceeds a predetermined threshold.

The invention discloses a voice retrieval method. The voice is voice data separated from video and audio data. The method comprises the following steps of: presetting: presetting an XML object database of material files or program files, wherein the XML object comprises XML element data for describing audio data and corresponding text attributes; and acquiring a voice characteristic quantity and a time code of each voice data, and associating each text character and the characteristic quantity and the time code of the corresponding voice respectively; searching: searching the matched text information in the XML object database according to the searching keywords submitted by a user, and extracting the corresponding audio information according to the voice characteristic quantity and the time code associated with the text; and outputting: displaying the video and audio information and the text information on a search result interface. The method is convenient for the user to simply and quickly acquire the desired video and associated text information, and does not occupy excessive system resources.

A data processing apparatus includes a speech recognition unit configured to perform continuous speech recognition on speech data, a related word acquiring unit configured to acquire a word related to at least one word obtained through the continuous speech recognition as a related word that is related to content corresponding to content data including the speech data, and a speech retrieval unit configured to retrieve an utterance of the related word from the speech data so as to acquire the related word whose utterance has been retrieved as metadata for the content.

The invention discloses an electronic card system which comprises a speech acquisition module, a speech analysis module, a speech characteristic extraction module, a keyword recognition module, an electronic card manager and an electronic card database, wherein the speech acquisition module is used for acquiring speech; the speech analysis module is used for transmitting an effective speech signal to the speech characteristic extraction module; the speech characteristic extraction module is used for extracting speech characteristics according to the received speech signal; the keyword recognition module is used for carrying out keyword recognition on the speech characteristics and outputting a keyword; the electronic card manager is used for establishing an electronic card in the electronic card database according to the keyword or browsing, retrieving, deleting and correcting the electronic card established in the electric card database; and the electronic card database stores the electronic card. The invention also discloses a speech recording method and a speech retrieval method of the electronic card, and also discloses an effective speech judgment method. According to the invention, contact information can be conveniently recorded and the electronic card is established, and multiple electronic cards can be rapidly retrieved.

A speech recognition support method in a system to retrieve a map in response to a user's input speech. The user's speech is recognized and a recognition result is obtained. If the recognition result represents a point on the map, a distance between the point and a base point on the map is calculated. The distance is decided to be above a threshold or not. If the distance is above the threshold, an inquiry to confirm whether the recognition result is correct is output to the user.

The invention discloses a speech retrieval method and system, and relates to the technical field of speech retrieval. The speech retrieval method comprises the steps that query speech is obtained; a second Hash sequence of the query speech is extracted; the second Hash sequence is matched with an established system Hash index table to obtain a first Hash sequence matched with the second Hash sequence; an original file is obtained in an established ciphertext speech library according to the first Hash sequence; the ciphertext speech library is established; and the system Hash index table is established. According to the speech retrieval method, a biological Hash technology is used for reference, feature extraction is carried out on original speech, the first Hash sequence of the original speech is obtained, the first Hash sequence is used as a retrieval abstract, and by comparing the Hamming distance between the first Hash sequence and the second Hash sequence of the query speech, matching of retrieval content is completed. By adopting the biological Hash technology, the speech Hash abstract is extracted, and the safety of the speech Hash abstract is improved.

The invention discloses a digital speech perception hash method based on formant frequency. The method is used for speech retrieval in a big data background, and the format frequency capable of reflecting timbre characteristics of speakers and time domain energy differences having the strong robustness can be respectively extracted to be used as the detail characteristics of the speech segments. During the matching process, the speech rough characteristics can be matched, and the speech segments having the timbres, which are similar to that of the target speech, can be screened out, and then the speeches having the similar timbres can be screened out for the matching of the detail characteristics, and at last, the accurate matching result can be acquired. When the method is used for the mass speech signal processing, a lot of unnecessary calculation amount can be saved, and the matching efficiency can be improved obviously.

The invention discloses a speech retrieval method and system based on fast Fourier inverse transformation. The method comprises the following steps: acquiring voice to be queried; carrying out featureextraction on the to-be-queried voice by adopting fast Fourier inverse transformation to obtain a to-be-queried feature vector; carrying out dimension reduction on the to-be-queried feature vector byadopting the measurement matrix to obtain a dimension-reduced to-be-queried feature vector; constructing a to-be-queried hash sequence according to the to-be-queried feature vector subjected to dimension reduction; matching the to-be-queried Hash sequence with a system Hash index table, and determining a matched Hash sequence; determining a retrieval voice file according to the matched hash sequence and the ciphertext voice library; and storing the system hash index table and the ciphertext voice library in the cloud server. According to the method, the fast Fourier inverse transformation iscombined with the measurement matrix, so that the speech features with good robustness and distinction can be efficiently extracted, and the retrieval efficiency and the retrieval accuracy are improved.

The invention relates to a voice retrieval method and system based on audio fingerprints. The method comprises the following steps: extracting Mel frequency cepstrum coefficient (MFCC) features and linear prediction cepstrum coefficient (LPCC) features of original voice with the duration of 20s; performing feature combination processing on the MFCC features and the LPCC features, and determining acombined feature matrix; performing column dimension reduction on the combined feature matrix based on an information entropy feature dimension reduction method, and determining a feature matrix after column dimension reduction; based on an energy-based feature dimension reduction method, performing row dimension reduction on the feature matrix after column dimension reduction, and determining afeature matrix after row dimension reduction; constructing an audio fingerprint database according to the feature matrix subjected to row dimension reduction; and performing matching retrieval on theto-be-queried voice segment and the audio fingerprint in the audio fingerprint library by using a normalized Hamming distance algorithm. According to the invention, the retrieval efficiency and retrieval precision of the long voice segment and the retrieval robustness of the audio fingerprint can be improved.

The invention relates to a speech deep hash learning method and system based on CNN. The method comprises the following steps: preprocessing an original voice file to obtain a preprocessed original voice file; extracting spectrogram features of the preprocessed original voice file; inputting the spectrogram features into an improved convolutional neural network model for training and deep hash feature learning to obtain deep semantic features of an original voice file; performing deep hash sequence construction on the deep semantic features by using a learned hash function to obtain a deep hash binary code representing the original voice file; and performing voice retrieval according to the deep hash binary code. According to the method, the problems of limitation, poor feature representation and the like of manual features in the feature extraction process of an existing content-based voice retrieval system can be solved, and the retrieval precision and the retrieval efficiency can befurther improved.

The invention discloses a commodity library retrieval system based on cloud inventory, and relates to the technical field of e-commerce management. The system comprises a commodity input unit, a commodity identification unit and a commodity retrieval unit. The commodity input unit comprises a commodity management module and a commodity storage module, the commodity recognition unit comprises a character recognition module, a voice recognition module, a picture recognition module, a preprocessing module and an output module, and the commodity retrieval unit comprises a commodity retrieval module, a user-defined sorting module and a commodity output module. According to the method, massive commodities are input into the cloud database, the input commodities are recognized and extracted to construct the feature library, the commodity feature library in the cloud inventory is matched in three modes of text retrieval, voice retrieval and picture retrieval, and matching results are arranged,so that the retrieval accuracy of customers is improved, and the retrieval efficiency of the customers is improved. Therefore, the purchasing time of customers is greatly reduced, and the purchasingdesire of customers is effectively increased.

The invention discloses a method and a device for trans-speech retrieval. The method comprises: receiving a case-retrieving request sent by a client; acquiring keywords carried by the case-retrieving request; and translating the keywords into a preset language and retrieving matched cases. Due to the method provided by the invention, foreign cases can be used for reference so as to ensure the comprehensiveness of retrieval results.

The present invention provides a content-based video retrieval method, comprising: receiving a video search request via a human-computer interaction interface; triggering a plurality of retrieval processes according to the video search request, including: triggering a metadata retrieval server to retrieve the metadata of a video program Retrieve; trigger the subtitle retrieval server to retrieve the XML file storing the subtitle text of the program; trigger the video retrieval cluster to retrieve the feature data of video key frames; trigger the voice retrieval cluster to search the voice information of the video program, including pinyin strings and pinyin diagrams Retrieve; integrate each search result, and then return it to the user through the human-computer interaction interface. This method provides an efficient solution for content-based video retrieval.

Disclosed are a speech retrieval apparatus and a speech retrieval method for searching, in an audio file database, for one or more target audio files by using one or more input search terms. The speech retrieval apparatus comprises a related document obtaining unit configured to search, in a related document database where documents related to audio files in the audio file database are stored, for one or more related documents by using the search terms; a correspondence audio file obtaining unit configured to search, in the audio file database, for one or more correspondence audio files corresponding to the obtained related documents; and a speech-to-speech search unit configured to search, in the audio file database, for the target audio files by using the obtained correspondence audio files.