Claims:We claim:
1. A method for real-time translation of speech terminal, comprising of
a) microphone, a speaker, and character information input and output modules,
b) a language selection module for sender or recipient voice and for text information;
c) a voice processing module to receive and process natural language selected by a user and / or text information for speech/text conversion
d) the semantic knowledge module to identify the natural language process voice and / or text message;
e) a control module syntax, semantics digital data translation correction process for the synthesis of natural language voice and / or text message during the syntax of natural language; transmitter,
f) Transmitting the encoded set of three-yuan semantic form of digital data, sent by way of a radio signal to the wireless network;
g) The receiver for receiving digital radio signals transmitted from other voice terminals and
h) the voice terminal comprising three yuan set encoding semantic database, a natural language voice library, Chinese phrase voice and a text Lexicon
2. The method as claimed in claim 1, wherein: said natural language voice speech database and store text in a voice terminal or a CPU of an electronic system, ROM and / or permanent memory.
3. The method as claimed in claim I, wherein: CPU module and the semantic recognition grammar is stored in the control module, a voice terminal or electronic system, ROM and / or the permanent memory in.
4. The method according to claim 3, wherein the method of operating a voice command, includes the steps of:
(1) select the category to be entered natural language;
(2) entered in step (I) selected natural language voice;
(3) the speech into three yuan set encoding digital data in the form of semantic processing module, the semantic knowledge module by the speech recognition, disambiguation, obtaining and storing semantic data correctly;
(4) one mapping semantic data corresponding to a voice command in the terminal or the electronic system sets;
(5) the voice signal in response to the program or execute a set operation of the electronic component.
5. The method as stated in claim 4, wherein the method for application synchronization with each other in real-time translation, includes the steps of:
(a) select the category to be entered natural language;
(b) entered in step (I) selected given natural language voice and / or text information;
(c) the speech processing module and / or text information is converted into a set of three yuan semantic digital data coded form by molding semantic cognition group and the voice / text message or cognitive, disambiguation, for the correct semantics;
(d) radio signals transmitted form the three yuan set of semantic encoding digital data over the wireless network;
(e) receives the transmission from the 3 yuan semantic data sets encoded speech terminal or another electronic terminal;
(f) to select the output of the natural language categories displayed;
(g) by semantic processing module 3 yuan encoded data set to be displayed translational synthesis voice and / or text information on the natural language grammar correction processing and to achieve the translation process by the natural language grammar control module; natural language voice display and / or a text message; and
(h) to the selected output.
6. The method as stated in claim 5, wherein the use voice and / or text message translation method according to any natural language, including the steps of: (1) voice terminal and / or multilingual translation program module, the speech database comprising natural language and / or text Lexicon, and grammatical control module;
(2) including any one of said pre-established and / or multilingual translation program mode group in the voice terminal; and
(3) any one of the downloading and / or programming languages translation module into the terminal through a wired or wireless voice mode;
(4) 3 yuan diversity receiving voice transmitted from another terminal semantic data in coded form;
(5) translation and / or more natural language voice and / or text message.
7. The method according to claim 6, wherein the method for application synchronization live video in real time online translation services, characterized by comprising a set of semantic data exchange 3 yuan encoded by wired or wireless manner, to provide real translators online translation services.
, Description:Technical Field of the Invention
The present invention relates to a method of fragmentation for multi-language subtitling and translation.
Background of the Invention
The Video fragmentation is a technology development in recent years gradually. The video player is a video sharing site launched a feature- based video sharing fragmentation, i.e., the user can use the tools provided by the site to intercept want to share a video clip from a video file and share it, but only limited to this. For example document CN102185880A discloses a video broadcast method and system for load balancing, terminal a request to play the video file to the root server debris; request to play a video file of a root server receives fragments sent from the terminal, the terminal to find the nearest server address margin, heavy orientation of the terminal connection address marginal server address; terminal requests a video file to a marginal fragmented server; if there video server marginal fragmented file, the server returns the marginal video file fragments to the terminal; if no video server marginal fragmented file, the marginal the server requests the registration of a live video source server fragmented file, a live video source server returns the file to the terminal pieces. And foreign viki Although there are other sites online translation subtitles, but the process of translation, audio and video files were not fragmented, when the translator Translator convenience is not good enough.
For the field of translation, one longer than required audio and video files in different languages dedicated translators take a long time, fragmentation of video can be more effective use of fragmented time translators to speed up the audio and video files subtitles progress of the translation, and the same videoclip can draw more than translators translation suggestions to enhance the quality of translation. At this point, the system automatically realizes how reasonable the audio and video files to fragment into a bottleneck.
Object of the Invention
The primary object of the present invention is to provide a method for multi-language subtitling
Another object of the present invention is to provide a method for multi-language subtitling implemented by using SNS community of users’ multi-language translation for one or more pieces of audio and video files;
Another object of the present invention is to provide a method for fragmented time translators
Summary of the Invention
The present invention solves the technical problem of the prior art subtitling can only be done by a group of independent subtitles, not many people realize synergies translation, and the large cost of translation, translation speed is slow, and there are people translated copyright and contract world the problem, providing a new way to implement multilingual subtitle translation. The method can reasonably be automatically audio and video file fragmentation and generate the corresponding audio and video file fragmentation, achieve in the shortest time at the lowest cost to complete the best multilingual subtitling.
Brief Description of the Drawings
FIG. 1 of the present invention based on multilingual subtitles art audio and video fragmented schematic implementation translation process.
FIG 2 audio and video files fragmented schematic diagram of the present invention.
[For purposes of the present invention, technical solutions and advantages clearer, explicit, the following detailed description of the present invention further. The specific embodiments described herein are only intended to illustrate the present invention and are not intended to limit the present invention.
Detailed Description of Invention:
A method for multi-language subtitling implemented, according to the properties of the original audio voice or subtitle of the video file which is fragmented into a plurality of audio and video files by the time dimension and to distribute audio and video files fragmented SNS community; the SNS is a social networking services; The SNS community of users multi-language translation for one or more pieces of audio and video files; It Select the audio and video of each fragmented file for each language translation of the best content and integrated into the corresponding language subtitles.
By analyzing the original audio and video files of subtitle or voice by analyzing the properties of audio and video files; The audio and video files to be parsed and exploded fragmented debris generated audio and video files. It has released the fragmented pieces of audio and video files to the SNS community; and SNS user community to translate segments of interest; The back office staff to collect the best translation for each language of each piece of audio and video clips and generate the best translation for each language subtitle file integration debris same language.
In the above aspect, preferably, the decomposition time dimension includes a start time point prior to entering a continuous or non-speech segment speech segment mute section as a sentence to the end of the last time continuous speech voice segment as the end point of the period of time of the sentence. The time dimension exploded comprising segments muting audio segment or speech segment is set to the minimum length of a pause and comprises 5 audio segments comprising voice segment length to the minimum sentence. The time dimension exploded comprising audio segment comprises voice segment 30 as the maximum length of the sentence. The said parsing comprises audio and video in original subtitle file: determining whether the original subtitle text subtitle file format, and if so, parsing caption file; otherwise, the text subtitle file into word format after parsing the subtitle file. To determining whether none of the original audio part of video files in the video file after the subtitle peeling generating an audio file, an audio file and then parsing. The parsing comprises an audio file different scene, distinguish voice conversation speech recognition process according to the voice and other sound properties of the audio and video files.
The present invention will be fragmented translation technique is applied to the multilingual subtitles, subtitle analytical method by way of the original speech analysis method and the combination of audio and video files is parsed and fragmentation. Specifically, for audio and video files have the original subtitles, stripped subtitle file and parses, audio and video files are fragmented split subtitle files according to the analysis result; for audio and video files without the original subtitles, video files need to strip after the sound analysis section generates the audio file (audio file is analyzed directly), analyzing the audio file, each segment obtained system time point in the audio and video files, and accordingly subjected to fragmentation to generate fragments divided file. Method of the present invention can automatically reasonable audio and video file fragmentation and generate the corresponding audio and video file fragmentation, achieve in the shortest time at the lowest cost to complete the best multilingual subtitling, and achieved good technology effect.
FIG. 1 of the present invention based on multilingual subtitles fragmented audio and video technology to achieve translation flowchart of a method, as shown in Fig, the method employs the steps of: I) users to upload audio and video files;
2) by analyzing the original audio and video files of subtitle or audio and video files according to attributes of the speech, and audio and video files is parsed exploded fragmented files and generates video and audio;
3) to distribute audio and video files fragmented SNS community;
4) SNS community of users involved in translation of audio and video files are fragmented, i.e., multi-language translation for one or more pieces of audio and video files;
5) select each fragmented file for each language best translations;
6) the integration of fragmented translations generates each language subtitles.
wherein the audio and video analysis and the subtitle file fragmentation process comprising: converting the vast majority of known srt subtitle file into subtitle file, parses; Further, audio and video files based subtitle free from video file to extract audio files in different scenes, dialogue distinguish sounds and other sounds, the analysis based on a specific voice, and accordingly obtain audio and video file fragmentation partitioning scheme. For example, can be carried out according to different speakers pronounce their unique characteristics of speech recognition based on a specific person, and accordingly obtain audio and video file fragmentation partitioning scheme.
The above process will be decomposed by the time dimension into a plurality of audio and video files audio and video files fragmentation process, is effective to achieve effective control of the length and number of sentences, to avoid very short will cause a lot of long sentence or several sentences occurs , respectively, the minimum pause length, the maximum length of the sentence and the sentence length minimum restrictions, while using the time dimension of decomposition. The minimum length limit pause role is to ignore short audio information, such as the speaker's instantaneous ventilation so as to ensure the integrity of the sentence. In practice, two audio segments to set the minimum length of a pause, i.e., a single non-continuous speech unit of the speech unit would not be considered a pause. Sentences of minimum length limit function is invalid filtered out short audio message, such as the speaker's cough, in practical applications, audio segments 5 is set to the minimum length of the sentence. The maximum length of the role of the sentence is too long to avoid a sentence unit, this method makes the sentence end soon, in practical applications, an audio segment 30 is set the maximum length of the sentence.
The above process generates subtitles in various languages in the display, for the convenience of viewers subtitles, results in longer sentences may be divided into multiple lines.
FIG 2 audio and video files fragmented schematic diagram of the present invention, as shown, it includes the following procedure:
I) users to upload audio and video files;
2) for analysis of audio and video files;
3) determines whether the audio and video file containing the original subtitles;
4) If the original audio and video files containing subtitles, it is further determined whether the audio and video file format srt; NO shell U, the audio portion of a video file by peeling generate an audio file;
5) If the original audio and video files contain subtitles for srt format, Yes, then press the original subtitle audio and video files are parsed; otherwise it will contain the original subtitle of audio and video files to srt format, and then press the original audio and video files subtitle parsing;
6) of the audio video file release part obtained by analyzing a voice file audio attribute of audio and video files;
7) to select the appropriate attributes to generate audio and video files audio and video files is parsed according to the segmentation scheme fragmentation;
8) to generate audio and video files audio and video files according to file fragmentation scheme.
The fragmentation process audio and video file comprises a video file to extract the audio stream by the audio classifier to distinguish the audio portion and the non-speech portion of the speech, the speech content of the fly-free speech portions separated off, and part of speech is segmented into a number of "sentence type" cell. For the audio classification system may classify selects between a Gaussian mixture model (Gaussian Mixture Model GMM), K- neighbours (K- Nearest Neighbour, KNN) , or a support vector machine (Support vector machine ,SVM), you may be provided such as a frame length, segments length and other configuration parameters.