Disclosure of Invention
The invention aims at the defects and provides an intelligent weather forecast system and a weather forecast voice segmentation method for a broadcasting station.
In order to achieve the above purpose, the present invention provides the following technical solutions: the intelligent weather forecast system of the broadcasting station comprises a recording unit, a recording unit and a control unit, wherein the recording unit is used for recording according to word statement data to form voice data;
the voice segmentation unit is used for converting the text sentence data into phonetic sentence data, and enabling the phonetic sentence data to correspond to the voice data one by one, so that corresponding initial consonant voice fragments and corresponding final voice fragments are found in the voice data by the initial consonant and the final in the phonetic sentence data, and the frequency of the initial consonant voice fragments is analyzed;
the voice segmentation unit comprises a frequency segmentation unit and is used for segmenting the rear ends of the initial consonant voice fragments with dense frequencies to form voice materials according to analysis results, and segmenting the pinyin sentences according to the voice materials to form corresponding pinyin sentence materials;
the voice database is used for storing the voice materials, the pinyin sentence materials, and music materials such as head music, tail music, background music and the like;
the information reading unit is used for reading weather data information through the Internet;
the information analysis unit is used for analyzing the weather data information of the information reading unit to form read pinyin sentence data, and splitting the read pinyin sentence data into pinyin sentence materials which can be identified by the voice database;
and the voice synthesis unit is used for finding the voice materials corresponding to each other from the voice database according to the phonetic sentence materials of the information analysis unit to synthesize the voice materials to form play voice.
Preferably, the voice segmentation unit further comprises a mute segmentation unit, which is used for segmenting according to the voice segments with the volume below 20 db as nodes.
Preferably, the voice synthesis unit further comprises a sound mixing unit, and the sound mixing unit is used for synthesizing the played voice after the front end of the played voice is added with the head music, the rear end of the played voice is added with the tail music and the background music, so as to form the forecast voice.
Preferably, the sound box speech segments with dense frequencies in the speech segmentation unit are sound box speech segments with frequencies above 8000 hz.
A weather forecast voice segmentation method is applicable to a broadcasting station intelligent weather forecast system, and comprises the following steps:
step S101, recording according to the required text sentence data to form voice data;
step S102, converting the text sentence data into phonetic sentence data;
step S103, finding out corresponding phonetic voice fragments in the voice data according to the phonetic in the phonetic sentence data;
step S104, finding out corresponding initial consonant voice fragments and final sound fragments in the pinyin voice fragments according to the initial consonants and final sounds in the pinyin;
step S105, when the frequency of the initial consonant voice segment is above 8000 Hz, cutting by taking the rear end of the initial consonant voice segment as a node to form a corresponding voice material;
step S106, the phonetic sentence data is segmented according to the voice material to form corresponding phonetic sentence materials;
step S107, the phonetic sentence material and the corresponding phonetic material are stored in the phonetic database.
Preferably, in step S105, the voice with the voice volume below 20 db in the voice material is taken as a node, and the voice material of the node is segmented to form the corresponding voice material.
Compared with the prior art, the invention has the following beneficial effects:
according to the intelligent weather forecast system of the broadcasting station, the word statement data to be played is recorded through the recording unit to form voice data, and the voice segmentation unit segments the voice data into voice materials to be stored in the voice database, so that the voice materials in the voice database are updated in real time; the weather data information is acquired on the Internet through the information reading unit, the information analysis unit analyzes the weather data information acquired by the information reading unit to form pinyin statement materials which can be identified by the voice database, and the voice synthesis unit analyzes the pinyin statement materials from the information analysis unit to find out corresponding voice materials in the voice database for synthesis, so that the weather forecast voice broadcasting can be played in real time, the data is accurate and timely, errors of manual voice broadcasting are reduced, meanwhile, the time of manual voice broadcasting is reduced, and the labor cost is saved; the sound mixing unit adds the head music, the tail music and the background music into the played voice, so that the played voice is milder and more comfortable in the listening process;
according to the weather forecast voice segmentation method, the voice data is segmented by adopting the mode that the frequency of the initial consonant voice segments is more than 8000 Hz as the segmentation node, so that voice is softer and not stiff when the segmented voice material is synthesized, the phenomenon of clamping and uncoordinated voice mutation is avoided, a better pronunciation effect is achieved, and the method is closer to real-time manual playing of personnel.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
The technical schemes of the embodiments of the invention can be combined, and the technical features of the embodiments can also be combined to form a new technical scheme.
Referring to fig. 1, the present invention provides the following technical solutions: the intelligent weather forecast system of the broadcasting station comprises a recording unit 1, a voice recording unit and a data processing unit, wherein the recording unit is used for manually recording according to word statement data to form voice data, and the word statement data is Chinese statement;
the voice segmentation unit 2 is used for converting the text sentence data into phonetic sentence data, and enabling the phonetic sentence data to correspond to the voice data one by one, so that corresponding initial consonant voice fragments and corresponding final vowel voice fragments are found in the voice data by the initial consonant and the final vowel in the phonetic sentence data, and analyzing the frequency of the initial consonant voice fragments;
the voice segmentation unit 2 comprises a frequency segmentation unit 21, which is used for segmenting the rear ends of the initial consonant voice fragments with dense frequencies to form voice materials according to analysis results, and segmenting the pinyin sentences according to the voice materials to form corresponding pinyin sentence materials;
a voice database 3 for storing the voice material and the pinyin sentence material, and music materials such as head music, tail music, background music, etc.; the phonetic materials comprise common phonetic materials and change phonetic materials, and the pinyin sentence materials comprise common pinyin sentence materials and change pinyin sentence materials;
an information reading unit 4 for reading weather data information through the internet; the method comprises the steps of obtaining real-time weather data of the county, the city and the county of the country from an official data organization such as a weather bureau and the like, wherein the real-time weather data comprise real-time temperature, highest and lowest temperature, real-time humidity, real-time air quality, real-time weather state, predicted weather change within 24 hours and the like;
the information analysis unit 5 is used for analyzing the weather data information of the information reading unit to form read pinyin sentence data, and splitting the read pinyin sentence data into pinyin sentence materials which can be identified by the voice database 3;
and the voice synthesis unit 6 is used for finding the voice materials corresponding to one from the voice database according to the phonetic sentence materials of the information analysis unit to synthesize the voice materials so as to form play voice.
The voice segmentation unit 2 further includes a silence segmentation unit 22, which is configured to segment (e.g. pause in a sentence, and corresponding to comma, colon, etc. in text sentence data) according to a voice segment with a volume below 20 db as a node. The voice segmentation unit 2 further includes a selection unit 23, configured to divide the pinyin sentence data into a common pinyin sentence portion and a variable pinyin sentence portion according to different application occasions, and for the voice data including the common pinyin sentence portion and the variable pinyin portion, the silence segmentation unit 21 is preferably selected to segment the voice data, and segment the voice data portion corresponding to the common pinyin sentence portion into common voice materials, and for the common voice materials, the common voice materials may not be segmented any more and are directly stored in the voice database 3 as voice materials; the speech data portion corresponding to the changed pinyin portion is split by the frequency splitting unit 21.
The speech synthesis unit 6 further includes a mixing unit 61 for synthesizing the played speech after the front end of the played speech is added with the head music, the rear end of the played speech is added with the tail music and the background music, so as to form a predicted speech. The speech synthesis unit 6 further comprises a breathing sound unit for adding breathing sound at the silence segmentation unit 21 in the synthesized playing speech when the number of initials and finals in the playing speech is greater than 40. The breath sounds are stored in the speech database 3.
The sound source sound fragments with dense frequency in the sound segmentation unit 2 are sound source sound fragments with frequency more than 8000 Hz. The frequency of the initial consonant voice segment is more than 8000 Hz, and the initial consonant is s, sh, q, x.
Referring to fig. 2, a method for voice segmentation of weather forecast is applicable to a broadcasting station intelligent weather forecast system, and a voice segmentation unit 2 in the broadcasting station intelligent weather forecast system processes voice data by adopting the method, the method comprises:
step S101, recording according to the required text sentence data to form voice data; the text sentence data is a Chinese character sentence; the recording is artificial recording, and different people record voice data with different tone colors;
step S102, converting the text sentence data into phonetic sentence data;
step S103, finding out corresponding phonetic voice fragments in the voice data according to the phonetic in the phonetic sentence data;
step S104, finding out corresponding initial consonant voice fragments and final sound fragments in the pinyin voice fragments according to the initial consonants and final sounds in the pinyin;
step S105, when the frequency of the initial consonant voice segment is above 8000 Hz, cutting by taking the rear end of the initial consonant voice segment as a node to form a corresponding voice material; the frequency of the initial consonant voice segment is 8000 Hz or more, and the initial consonant is s, sh, q, x;
step S106, the phonetic sentence data is segmented according to the voice material to form corresponding phonetic sentence materials;
step S107, a voice database 3 is established, and the phonetic sentence materials and the corresponding voice materials are stored in the voice database 3.
The step S105 further includes using the voice with the voice volume below 20 db in the voice data as a node, and splitting the voice data of the node to form the corresponding voice material.
In the step S105, the pinyin voice segment may be further divided into a common pinyin voice segment and a variable pinyin voice segment by using the voice with the voice volume below 20 db in the voice data as a node, and dividing the voice data of the node; and secondly, detecting and judging the varied pinyin voice segment, and when the frequency of the initial consonant voice segment is more than 8000 Hz, cutting by taking the rear end of the initial consonant voice segment as a node to form a corresponding voice material.
Example 1:
the working principle of the intelligent weather forecast system of the broadcasting station in the invention is as follows:
recording the text sentence data (such as' the current air temperature is sixteen ℃ C.) commonly used in weather forecast through a recording unit 1 to form voice data; the speech segmentation unit 2 converts the text sentence data into pinyin sentence data (e.g. "d ā n E.G. qi n q. Mu.w n w. Mu.i: the sh I is the sh d), the selection unit 23 divides the phonetic sentence data into a common phonetic sentence part (such as'd ā n-size qia n q-size w n w ei'), a changed phonetic sentence part (such as 'sh I is the sh d is the d), for the phonetic data comprising the common phonetic sentence part and the changed phonetic sentence part, the mute segmentation unit 21 is preferentially selected to segment the phonetic data (short pause occurs in the sentence at the colon), the phonetic data part corresponding to the common phonetic sentence part is segmented into the common phonetic material, the common phonetic material can be directly stored in the phonetic database 3 as the phonetic material, the common phonetic sentence part is stored in the common phonetic sentence part (such as' sh I is the d 'and the common phonetic sentence part), the common phonetic sentence material is used for encoding the common phonetic material (the Chinese character mode (such as' current) or the Chinese character and the vowel addition unit 6 is convenient to query;
when the frequency of the initial sound fragment is above 8000 Hz through the frequency segmentation unit 21, the back end of the initial sound fragment is used as a node to segment the sound data corresponding to the changed phonetic sentence part (such as 'sh', 'I Li' and 'E' and 'D' and the changed phonetic sentence part corresponding to the segmented changed phonetic sentence part is stored in the sound database 3, and the changed phonetic sentence material is used for encoding the changed phonetic material (the encoding can also adopt a Chinese character mode or a Chinese character and initial consonant adding mode (such as 'I six' and) and is convenient for the inquiry of the sound synthesis unit 6;
the information reading unit 4 obtains real-time weather data information (for example, "the current air temperature is 16 ℃) through the Internet, and transmits the weather data to the information analyzing unit 5; the information analysis unit 5 analyzes the weather data information to form read pinyin sentence data (such as'd ā n Gama' n q n Gama 'n w Ei: sh's sh'd' and split the read pinyin sentence data into pinyin sentence materials (such as'd ā n' qi n q Ei's n w Ei', 'sh' and 'i's sh'd' and split the read pinyin sentence data into varied speech materials with frequencies of more than 8000 hz voice fragments if the split read pinyin sentence data contains a common pinyin sentence part and a varied pinyin sentence part, and split the varied pinyin sentence part into varied speech materials by the information analysis unit 5;
the voice synthesis unit 6 finds out one-to-one corresponding common voice materials and changing voice materials in the voice database 3 according to the common pinyin sentence materials and the changing pinyin sentence materials analyzed by the information analysis unit 5, synthesizes the common voice materials and the changing voice materials, and forms play voice; the mixing unit 61 finds out the corresponding piece of head music, piece of tail music and background music from the voice database 3, loads the corresponding positions of the played voices and synthesizes the corresponding positions to form forecast voices for playing.
The whole process of the intelligent weather forecast system of the broadcasting station acquires data through connecting with the Internet, the data are accurate and timely, manual operation is not needed, the computer can automatically complete the intelligent weather forecast system, the synthesis time is not more than 1 second, finally, the synthesized voice comprises real-time weather data of all places of the whole country and weather changes within 24 hours, weather early warning and warmth prompt are realized, the broadcasting station is still available in non-working time and legal holidays, and the time and labor cost are saved.
The working principle of the weather forecast voice segmentation method in the invention is as follows:
in step S101, manual recording is performed according to the required text sentence data (such as "the current air temperature is sixteen ℃), so as to form voice data, and then step S102 is performed;
in step S102, the text sentence data is converted into phonetic sentence data (e.g. "d ā n site qi n q. Mu.w n w. Mu.i: sh I sh d. Mu."), followed by step S103;
in step S103, corresponding pinyin voice segments are found in the voice data according to the pinyin in the pinyin sentence data, each pinyin is found to correspond to one of the pinyin voice segments, and step S104 is performed;
in step S104, corresponding initial consonant speech segments and final sound segments are found in the pinyin speech segments according to the initial consonants and final sounds in the pinyin, and then step S105 is performed;
in step S105, when the frequency of the initial consonant voice segment is above 8000 hz, the back end of the initial consonant voice segment is used as a node to segment the voice data, the first voice data (such as "d ā n #," i_nq "," i_w_nwysish "," i_li_sh "," i_w_h "", and the second voice data (such as "d ā n_q", "i_nq", "i_w_nwi", "sh_h) formed after segmentation is used as a node, and the second voice data (such as" d ā n_q "," i_w_nq "," sh "," i_li_d) formed after segmentation is further used as a voice material, and the second voice data segment after segmentation is used as a voice material;
in step S106, the phonetic sentence data is segmented according to the phonetic material, forming corresponding Pinyin sentence materials (e.g. "d ā n q", "i nq", "i w nwe ish", "i liu", "e sh", "i d); cutting the phonetic sentence data based on the cutting point of the phonetic material, and then performing step S107;
in step S107, a speech database 3 is established, and pinyin sentence materials and corresponding speech materials are stored in the speech database; wherein the phonetic sentence material phonetic material corresponds to the code or code number in the phonetic database 3.
In another embodiment of the weather forecast voice segmentation method, the step S105 is improved, the voice with the voice volume below 20 db in the voice data is divided into nodes, the voice data of the nodes is segmented, the first voice data (such as'd ā n, qi n q μw n w ei', 'sh i li'd ') formed after segmentation is directly used as the voice material, and the first voice data is arranged according to different use environments to form common pinyin voice segments (such as'd ā n, qi n q a, w n w i ') and variable pinyin voice segments (' sh i'd') which are directly stored in the voice database 3; secondly, detecting and judging the varied pinyin voice segment, when the frequency of the initial consonant voice segment is more than 8000 Hz, cutting the back end of the initial consonant voice segment as a node, and forming second voice data (such as ' sh ', ' Li ' sh ', ' e sh ','d ' after cutting) which are formed after cutting, thereby forming corresponding voice materials and storing the voice materials in the voice database 3.
When synthesizing the voice materials in the voice database 3, firstly converting the word sentence data to be synthesized into pinyin sentence data, and then splitting the pinyin sentence data by taking the initial consonants (such as s, sh, q, x and the like) with the voice fragment frequency of more than 8000 Hz and the symbols (such as commas, colon and the like) as nodes to form the pinyin sentence materials; the corresponding voice materials are found in the voice database 3 through the phonetic sentence materials to be connected and synthesized, and play voice can be formed; the voice material formed after the voice data is segmented by taking the rear end of the initial consonant voice segment as a node and voice with voice volume below 20 dB in the voice data as a node, when the voice material is spliced and synthesized to play voice, the voice sounds softer and not stiff, the phenomenon of clamping and voice mutation incompatibility can not occur, the voice playing device has a better pronunciation effect, and is closer to real-time manual playing of personnel.
Although embodiments of the present invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made therein without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims and their equivalents.