JP2014170135A - Outdoor environmental sound transmitting device, and outdoor environmental sound transmitting system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an outdoor environmental sound transmitting device for transmitting a sound message that is easy to hear for local residents outdoors in an echo environment.SOLUTION: An outdoor environmental sound transmitting device 100 for transmitting a sound message for disaster prevention to one or more branch stations laid in an area comprises: message division means 33 for dividing the sound message into meaningful symbol strings or obtaining the sound message divided into the meaningful symbol strings; speech period calculating means 34 for calculating a speech period for each of the symbol strings; section position determining means for determining the end of one or more of the symbol strings, in which the total of the speech periods is a delay period or less, as a section position of the sound message transmitted to each of the branch stations; soundless character inserting means 35, 36 for inserting soundless characters for the delay period or more at the section position of the sound message; and message transmission means 21 for transmitting the sound message, into which the soundless characters have been inserted, to the branch stations.

Description

  The present invention relates to an outdoor environment voice transmission device that transmits a voice message for disaster prevention to one or more slave stations laid in an area.

  Various disasters are caused by natural phenomena and human factors. Governments and local organizations monitor the occurrence of disasters that could affect human lives, and appropriately evacuate local residents when disasters occur. Occurrence of disasters and evacuation requests are transmitted to local residents through television broadcasting, radio broadcasting, mobile communication, outdoor loudspeakers, and the like (see, for example, Patent Document 1).

  There may be a case where a voice message that conveys the contents of a disaster that occurred or the evacuation destination is accompanied as information to be transmitted. For example, in the earthquake bulletin, the voice message “Meteorological Agency emergency earthquake bulletin. Earthquake has occurred. Please ensure your safety.” The tsunami warning says “Tsunami damage will occur. A voice message saying “People along the river should evacuate to safe places such as hills and evacuation buildings immediately.” Is transmitted, and the tsunami warning says, “The sea and the coast are dangerous. People in the sea. You should get up from the sea and leave the coast immediately. " Local residents can quickly take appropriate actions by grasping the contents of voice messages.

  However, there is a problem that the voice message by the outdoor loudspeaker, which is one of the transmission methods, may not be heard by the local resident depending on the local resident environment. For example, when a local resident stays indoors and does not watch TV, the voice message is blocked, and the local resident cannot grasp the content of the voice message. For local residents in such a situation, a technique for supplementing the transmission of voice messages by an outdoor loudspeaker has been conventionally devised (see, for example, Patent Document 2). Patent Document 2 discloses a disaster prevention alarm linkage system in which a home guard installed in a house communicates with a master unit to notify an inhabitant in the house of an earthquake early warning or the like.

JP 2001-273455 A JP 2013-029944 A

  However, even when the local residents are outdoors, there is a problem that the local residents may not be able to hear the voice message. In other words, the loud sound produced by the outdoor loudspeaker is loud and can be transmitted far away. On the other hand, depending on the geographical features of the area, an echo (Yamahiko, reverberation sound) is generated, making it difficult for local residents to hear. there were.

  In view of the above problems, an object of the present invention is to provide an outdoor environment audio transmission device that transmits an audio message that can be easily heard by local residents outside in an echo environment.

  The present invention is an outdoor environment voice transmission device that transmits a voice message for disaster prevention to one or more slave stations laid in an area, the voice message acquisition means for acquiring the voice message, and the voice message acquisition means Divides the voice message acquired by, into a meaningful symbol string, or message dividing means for requesting the voice message to be divided into the symbol string, and an utterance time for calculating the utterance time of each symbol string A calculating means; a delay time database in which a delay time of echo generated when the voice message is amplified from a slave station is registered for each slave station; and the one or more symbol strings whose total speech time is equal to or less than the delay time Each time, a process for determining the end of the symbol string as the voice message delimiter position is a voice message transmitted to the slave station registered in the delay time database. A delimiter position determining means for the message based on the delay time of the slave station, an unvoiced character inserting means for inserting an unvoiced character whose utterance time is longer than the delay time at the delimiter position of the voice message, and an unvoiced character is inserted. Message transmitting means for transmitting the voice message to the slave station.

  In an echo environment, it is possible to provide an outdoor environment audio transmission device that transmits an audio message that can be easily heard by local residents outside.

It is an example of the figure explaining the echo of a voice message. It is an example of the figure explaining the outline of a process of a disaster prevention radio | wireless system. It is a figure which shows the structural example of the whole mechanism of a disaster prevention radio | wireless. It is an example of the block diagram of a disaster prevention radio system. It is a figure which shows the experimental result of the speech intelligibility test by echo. It is an example of the hardware block diagram of a pose insertion apparatus. It is an example of the functional block diagram of a pose insertion apparatus. It is a figure which shows an example of an utterance time database and a delay time database. It is an example of the figure explaining the determination method of a delimiter position. It is an example of the figure explaining pause time. It is another example of the functional block diagram of a pose insertion apparatus. It is an example of the flowchart figure which shows the operation | movement procedure of a pose insertion apparatus. It is an example of the functional block diagram of a pose insertion apparatus (Example 2). It is a figure which shows an example of the delay time and echo / intelligibility correspondence database which a long path echo calculation part calculates. It is an example of the flowchart figure which shows the operation | movement procedure of a pose insertion apparatus.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. However, the technical scope of the present invention is not limited to this embodiment.

  First, the echo of a voice message will be described with reference to FIG. The slave station shown in the figure has a so-called speaker, and a voice message that conveys the contents of the disaster, an evacuation destination, etc. is louded out from the speaker. For example, when a speaker is directed to a local inhabited by a village, a voice message is transmitted directly from the speaker (hereinafter referred to as direct sound), but if there is a reflector such as a mountain, the same sound is reflected by the reflector. A message (hereinafter referred to as “reflected sound”) is transmitted to the local residents over time.

Since the direct sound and the reflected sound have different transmission paths, the same voice message is transmitted to the same local residents after the delay time Δt has elapsed. Assuming that the linear distance between the slave station and the local residents is L ₀ , and the route from the slave station to the local residents via the reflector is L ₁ + L ₂ , the delay time Δt is “L ₁ + L ₂ −L ₀ ” at the speed of sound. The divided value. In the example shown in the figure, the direct sound “please” and the reflected sound “evacuate to the hill” are heard, making it difficult to hear (in this way the voice message is transmitted). This will be referred to as delay overlap transmission).

  Therefore, the disaster prevention wireless system of the present embodiment suppresses delay overlap transmission by inserting a pause in the voice message.

FIG. 2 is an example of a diagram illustrating an outline of processing of the disaster prevention wireless system according to the present embodiment.
The upper part of FIG. 2A shows an example in which a voice message is reflected by an echo but does not result in delayed overlap transmission. If the sentence length of “Please evacuate to the hill” is Ts [ms], if the delay time Δt is longer than the sentence length, the reflected sound will be transmitted after the direct sound transmission is completed, so the delay overlap transmission will occur do not do.

However, in reality, it is rare that the delay time Δt is longer than the sentence length, so that delay overlap transmission occurs as shown in the lower part of FIG. The disaster prevention radio system according to the present embodiment suppresses the occurrence of delayed overlap transmission as follows.
(i) Divide voice message into phrases with delay time Δt or less
(ii) Inserting a Pause of Δt or More Between Phrases FIG. 2B is an example of a diagram for explaining the division of the voice message. The voice messages are divided into “Upper hill”, “Evacuate”, and “Please” so that each phrase is less than the delay time Δt. Although the phrase lengths of the phrases are strictly different, here, a common phrase length Ps [ms] is used. Therefore, Ps ≦ Δt.

  FIG. 2C is an example of a diagram schematically showing the insertion of a pose between phrases. A pause of Δt time is inserted after “on the hill”. Since Ps ≦ Δt, the next direct sound “evacuate” is not transmitted to the local residents until the transmission of the reflected sound “on the hill” is completed. Therefore, the local residents can hear the reflected sound “on the hill” after the direct sound “on the hill”, and the direct sound and the reflected sound are not transmitted in a delayed manner.

  When the first pause time (silent time) elapses, the direct sound “evacuate” is amplified and a pause time of Δt or more is inserted. During this pause time, the reflected sound “evacuate” is transmitted. Therefore, the direct sound and the reflected sound are not transmitted with delay overlap. Further, when the second pause time elapses, the direct sound “please” is amplified and the reflected sound “please” is transmitted following the direct sound “please”.

  Therefore, by inserting a pause of appropriate time between phrases, the situation in the upper part of FIG. 2A can be created arbitrarily, and the occurrence of delayed duplicate transmission is suppressed for any voice message and delay time Δt. it can. As will be described later, even if the direct sound and the reflected sound slightly overlap, the intelligibility of the voice message is not greatly affected, so the end of the direct sound and the beginning of the reflected sound overlap by about one to several characters. May be. That is, it is not necessary to completely eliminate duplication.

  FIG. 3 is a diagram illustrating a configuration example of an overall mechanism of the disaster prevention radio. Such a mechanism is called an outdoor environment audio transmission system or a disaster prevention radio system. The national government has established a system for transmitting information to the local governments, such as municipalities, on information that should be dealt with immediately, such as those obtained by the Japan Meteorological Agency or the Fire and Disaster Management Agency. National and local governments can communicate via the Internet and satellite communications. Information detected and transmitted by the Japan Meteorological Agency includes emergency earthquake warnings, tsunami warnings, tsunami warnings, eruption warnings, eruption forecasts, Tokai earthquake prediction information, seismic intensity warnings, weather information, landslide disaster warning information, tornado warning information, record There are short-term heavy rain information, designated river flood forecasts, national protection information such as terrorism and missiles. Hereinafter, these are referred to as “emergency information” without distinction.

  The local public entity has a receiving antenna 102 for receiving radio waves from the artificial satellite 101 and a communication device for connecting to the Internet (a network 24 described later), and can always receive emergency information without any human intervention. .

  The local government receives a voice message transmitted from the Japan Meteorological Agency, etc., or selects a voice message appropriate for the situation, and transmits it to the broadcast slave station 200 (the mobile slave station is also the same). Sent). The slave station 200 is systematically laid in the area so that local residents can perceive voice messages by hearing without omission.

  The local public entity has a telemeter 300 such as a river water level measuring instrument or a seismometer. The data transmitted from the telemeter 300 is discriminated based on its own standard, and the voice message is amplified from the slave station 200. It is also possible.

  FIG. 4 shows an example of a block diagram of a disaster prevention radio system arranged in a local public entity. The disaster prevention radio system 100 is an example of an outdoor environment audio transmission device according to the claims. The console 20 receives an operation when a user (a disaster prevention person in a local public organization) transmits a voice message, and controls the voice message transmitted via the master station communication device 21. Further, the console 20 receives emergency information or a voice message transmitted by the Japan Meteorological Agency or the like via the network 24, and transmits a voice message via the master station communication device 21 without any user operation.

  The console 20 includes a microphone 11, a display unit 12, an operation unit 13, a data storage unit 14, a control unit 15, and an input / output I / F 16, and the control unit 15 is connected to the master station communication device 21. It is connected.

  The control unit 15 is an information processing apparatus such as a microcomputer provided with a CPU, RAM, ROM, I / O, and the like. The control unit 15 controls the entire console 20 by an OS (Operating System) or a program executed by the CPU.

  The microphone 11 collects the voice uttered by the user and converts it into an electrical signal. The electric signal is encoded by the control unit 15 by a predetermined method and converted into an audio signal. The encoding method is not particularly specified, but there are 26.6 kbps, 6.4 kbps, and the like as standard code acceleration. The audio signal is stored in the data storage unit 14.

  The display unit 12 is a flat panel display such as a liquid crystal, and is preferably equipped with a touch panel. In addition to the operation menu, the display unit 12 displays a map captured from the outside, a video taken by the camera, and the like.

  The operation unit 13 is a hard key such as a keyboard, a soft key displayed on the touch panel, and a pointing device. The operation unit 13 accepts selection of a voice message transmitted from the slave station 200, for example.

  The data storage unit 14 is an HDD (Hard Disk Drive) or an SDD (Solid State Drive), and stores voice signals, text data of voice messages, programs, and the like.

  The input / output I / F 16 is an interface for communicating with peripheral devices, and is, for example, an Ethernet (registered trademark) card, a USB host, or the like. The input / output I / F 16 is connected to a FAX device 22, a speech synthesizer 23, a pause insertion device 30, and a network 24.

  The FAX apparatus 22 receives image data from the telephone line network or the network 24 and prints it, or transmits a text requested to be transmitted from the console 20 to a designated destination. The speech synthesizer 23 is a device that converts text data into speech. By converting the voice message of the emergency information requested from the console 20 into voice, the voice signal can be transmitted to the slave station 200 without the user reading out the voice message. The voice created by the voice synthesizer 23 is converted into a voice signal by the voice synthesizer 23 or the control unit 15 and recorded in the data storage unit 14.

  The pause insertion device 30 is an information processing device such as a PC (Personal Computer) or a microcomputer. Although the function of the pose insertion device 30 will be described later, the function of the pose insertion device 30 may be configured to be included in the control unit 15.

  The network 24 is mainly a LAN, but is connected to an IP communication network such as the Internet via the LAN. For this reason, information from outside such as J-ALERT (National Instantaneous Warning System) can be received.

  The control unit 15 wirelessly transmits the audio signal stored in the data storage unit 14 from the master station communication device 21. In the disaster prevention wireless system 100, a wireless communication standard (for example, a municipal digital broadcast communication system standard ARIB STD-T86) is defined. That is, the transmission unit 17 modulates a voice signal with π / 4 shift QPSK, and converts one frame into six time slots by a TDMA-TDD (Time Division Multiple Access-Time Division Duplex) method. To divide. Except for the control slot, audio signals can be stored in five slots. The modulated audio signal is transmitted from the antenna 25 with a carrier of 60 MHz, for example.

  In the same standard, the transmitter 17 of the master station communication device 21 instructs the slave station 200 to start reporting using the paging channel. A frame flowing in the paging channel includes a prefecture code and a user code (specifying a slave station), and the master station communication device 21 sets an arbitrary user code, whereby a slave station (or group of slave stations) 200 is set. Can be called individually.

  Note that this communication standard is only an example of a communication method determined by a local public body, and the like, even if the master station communication device 21 and the slave station 200 communicate with each other via a mobile phone network, a PHS network, a WiMAX network, a wireless LAN network, or the like. Good. Further, a communication network in which slave stations communicate with each other like ZigBee (registered trademark) may be constructed, and the slave station 200 and the master station communication device 21 may communicate with each other. Moreover, you may employ | adopt wired communication in one part area or all the areas instead of a radio | wireless.

  In addition, the slave station 200 receives a voice message from the master station communication device 21 and broadcasts it to the local residents. The slave station 200 includes a receiving unit 51, a control unit 52, a microphone 54, a speaker 53, and the like. In addition, although not shown, an operation unit, a display unit, and the like may be included. Further, only the slave station 200 can broadcast its own station such as a pseudo siren sound and a chime sound.

  The receiving unit 51 of the slave station 200 demodulates the carrier wave received by the antenna 55, extracts an audio signal, and outputs it to the control unit 52. The control unit 52 converts the audio signal into a voltage, amplifies it, and outputs it from the speaker 53. The microphone 54 collects sounds around the slave station 200. It is also possible to transmit this audio signal from the slave station 200 to the master station communication device 21.

[Ease of recognizing words in an echo environment]
Applicants have experimented with the effectiveness of language control in an echo environment.

1. The voice of the speaker mya recorded in the FW07 database [Kondo et al., IEICE Tech. Bulletin, SP2007-157, 2008.] was used as the sound source for the experiment contents. For the word familiarity, the top two of the four ranks (high familiarity: 7.0-5.5, low familiarity: 5.5-4.0) were used. In this experiment, an environment with one or two subsequent sounds was simulated.

  FIG. 5A shows the time pattern of the subsequent sound generated by the echo. As shown in the figure, a quadruple word was used as one listening sound, and one or two subsequent sounds were added with a delay time specified there. ○ □ △ × in the figure represents four different words. Since one word is 4 mora and the length is about 750 [ms], the time shift of 1/2 word length is 375 [ms]. The subsequent sound used was the same sound as the original sound. Because of the nature of the FW07 database, each listener needed to listen to all 400 words of one familiarity condition, so five speech sets were created, and 80 speech words (20 quadruple words in each set). ) Was randomly distributed. There are 7 types of time lags of the following sounds (conditions A to G in FIG. 5A), and at least 7 listeners are required to avoid re-listening the already-listened quadruple words under different conditions. It was necessary. For this reason, five different listeners were assigned to each condition. Seven listeners participated in both high and low intimacy experiments. The listener was instructed to sit on the chair in the soundproof room and fill in the answer sheet on the answer sheet in 15 seconds after one quad word was presented. The sound was presented to both ears via headphones (SennheiserHDA-200).

2. Results and discussion
FIG.5 (b) is a figure which shows the average correct answer in each experiment condition of high and closeness. Black bars in the figure indicate high familiarity and white bars indicate low familiarity conditions. Two-factor analysis of variance was performed with seven time-shift conditions and two intimacy conditions as factors within the subject. As a result of the analysis, the main effect of the time lag condition (F (6,
594) = 141.63, p <.001), main effect of intimacy condition (F (1, 99) = 85.98, p <.001), interaction between time shift and intimacy (F (6, 594) = 2.15 , p <.05), both were significantly different. In addition, except for the condition of only the original sound, the correct answer rate when the familiarity was high was significantly high. This suggests that intimacy is an effective vocabulary difficulty control index even in outdoor environments where long-path echo is present. When the analysis was performed according to intimacy, there was no significant difference in the average accuracy rate (between E and G) when there were two subsequent sounds under the high familiarity condition, but there was only one subsequent sound. A significant difference was observed between the conditions in which the delay time of the subsequent sound was one word length (B and D). When the number of subsequent sounds is two, the accuracy rate is generally low, so it is considered that the influence of the delay time of the subsequent sounds was not observed. Although the same tendency was observed in the low familiarity condition, a significant difference was observed between the conditions in which the delay time of the subsequent sound was one word length even when the number of subsequent sounds was two (B and D, E and G). In the future, we plan to calculate the word intelligibility for each appearance position and conduct a detailed study.

5. Summary The effect of intimacy on word intelligibility in the presence of longpass echo was investigated. As a result of experiments, it became clear that word intelligibility is strongly influenced by echo.

[Pause insertion device]
FIG. 6 shows an example of a hardware configuration diagram of the pose insertion device 30. The pause insertion device 30 includes a CPU 301, a ROM 302, a RAM 303, an HDD 304, a graphic board 305 to which a display 320 is connected, a keyboard / mouse 306, a media drive 307, and a network communication unit 308 connected to a bus. The CPU 301 expands and executes the program 310 stored in the HDD 304 in the RAM 303, and controls each component to perform input / output and process data. The ROM 302 stores a BIOS and a start program for reading the bootstrap loader from the HDD 304 to the RAM 303. The bootstrap loader reads the OS from the HDD 304 to the RAM 303.

  The HDD 304 may be a non-volatile memory, and may be an SSD (Solid State Drive) or the like. The HDD 304 stores an OS, a device driver, and a program 310 that provides functions to be described later. The OS includes Windows (registered trademark), LINUX (registered trademark), UNIX (registered trademark), and the like. The display 320 displays a GUI screen generated by the graphic board 305 as instructed by the program.

  A keyboard / mouse 306 is an input device that receives user operations. Note that the keyboard / mouse 306 and the display may not be provided when the user does not directly operate the pose insertion device 30.

  The media drive 307 reads and writes data to and from optical media such as compact discs, DVDs, and Blu-ray discs. Further, data may be read from and written to a memory card such as a flash memory.

  The network communication unit 308 is, for example, an Ethernet (registered trademark) card for connecting to a LAN. The OS and program 310 perform TCP / IP (UDP / IP) and application layer protocol processing. There are various application layer protocols, but they correspond to general protocols (for example, HTTP, FTP, SNMP (Simple Network Management Protocol), etc.).

  The program 310 is a file in an installable or executable format, and is distributed by being recorded on a computer-readable recording medium. The program 310 is distributed in a file that can be installed or executed from a server (not shown).

  FIG. 7 shows an example of a functional block diagram of the pose insertion device 30. The audio signal stored in the data storage unit 14 of the console 20 is acquired by the audio signal acquisition unit 31 via the input / output I / F 16. Since the voice signal acquisition unit 31 has acquired the voice signal of the utterance “Please evacuate to the hill”, the voice recognition unit 32 converts it into text data. For the speech recognition unit 32, publicly known software may be used. Generally, recognition processing is performed using an acoustic model and a language model, and converted into text data. Further, a voice signal may be transmitted via the network 24 and voice recognition may be performed using an external server (for example, a cloud service). The voice recognition unit 32 sends the text data to the phrase division unit 33.

  The phrase dividing unit 33 divides the text data into phrases. A phrase is a symbol string of one or more characters that is meaningful to humans. For example, a phrase can be interpreted as a word. However, in terms of grammar, “ni”, “te”, etc. are also words (attached words), and character strings such as “on the hill” and “evacuate” are “words (independent words) + attached words”. On the other hand, if there is no accompanying word, it may be a phrase with only a word. Therefore, the phrase of the present embodiment is not limited to a word and may be a meaningful symbol string. Of course, numbers and alphabets may be included in addition to letters.

  The reason for dividing the phrase is to determine a delimiter position for inserting a pause. Since the phrase utterance time Ps is preferably equal to or less than the delay time Δt, the phrase is divided into phrases of the shortest symbol string that are meaningful (for example, including independent words).

  In a divided language such as English (a language in which sentences are divided for each word), it is divided into words when converted into text data. Therefore, it is only necessary to divide each word. However, as with Japanese, it is difficult for local residents to grasp the meaning of the “a” and “the” definite articles and indefinite articles. Combine words to make a phrase.

  In languages such as Japanese, which are not separated, it is preferable to divide them into words once to divide them into phrases. As a simple method of dividing into words, Kana-Kanji conversion is effective for languages that can be converted to Kanji such as Japanese (Chinese, Korean, Vietnamese, etc.). Kana-Kanji conversion can be performed by a general IME (Input Method Editor), and an external server can also be used.

  From the kana-kanji converted text data, the kanji to the next kanji are estimated as one word. For example, when Kana-Kanji conversion is performed and text data “Please evacuate to high ground” is obtained, “To high,” “Evacuate” and “Please” are the words.

  It is also effective to perform morphological analysis as a method of dividing Japanese into words. Morphological analysis is a process of dividing a sentence written in a natural language into the smallest units having grammatical meaning. In Japanese morphological analysis, it can be divided up to the part of speech level. In the text data “Please evacuate to high ground”, morphemes “high hill”, “ni”, “evacuate”, “te”, “down”, and “i” are extracted. The phrase dividing unit 33 treats the particle as one phrase together with the immediately preceding noun, verb, adjective, adjective verb, and the like. Note that an external server can also be used for phrase division.

  The text data divided into phrases in this way is sent to the utterance time length calculator 34. The utterance time length calculation unit 34 refers to the utterance time database 37 and calculates the utterance time of each phrase.

  FIG. 8A shows an example of the utterance time database 37. In the utterance time database 37, the number of mora of each phrase is registered. Mora is a unit of Japanese sound, and if the number of mora is determined, the utterance time of the entire phrase can be determined. For example, the standard utterance time is set to 750 [ms] at 4 mora. Since the number of mora of “to hill” is 5, the utterance time of “to hill” is “187.5 (1 mora utterance time) × 5 = 937.5 [ms]”.

  Note that the utterance time may be registered instead of the number of mora. In addition, the number of mora of words instead of phrases may be registered. When the number of mora of words is registered, the number of mora of each word included in the phrase is totaled.

  Returning to FIG. 7, the utterance time length calculation unit 34 attaches the utterance time to each phrase and sends it to the delimiter position / pause time determination unit 35. For example, in “Up to hill”, “Evacuate”, “Please”, the utterance time for each phrase is attached, such as “To hill: 937.5”, “Evacuate: 937.5”, “Please: 750”.

  The delimiter position / pause time determination unit 35 refers to the delay time database 38 and determines the delimiter position and pause time for inserting a pause.

  FIG. 8B shows an example of the delay time database 38. In the delay time database 38, a delay time Δt is registered in association with the slave station ID for identifying the slave station. For the slave stations with the same delay time, the delimiter position and pause time need only be determined once. For example, in the slave station with the slave station ID = 1, the delay time is 1000 [ms]. As described with reference to FIG. 2A, if the sentence length Ts of the entire voice message is equal to or shorter than the delay time, there is no need to insert a pause because no delay overlap transmission occurs. In addition, the time may be overlapped as long as the time does not affect the intelligibility of the voice. This time can be determined experimentally assuming that the decrease in the intelligibility of the voice is below a threshold value.

  When the sentence length Ts of the entire voice message is not equal to or shorter than the delay time, the break position / pause time determination unit 35 determines the break position and pause time according to the following procedure.

  FIG. 9 is an example of a diagram illustrating a method for determining a break position. As shown in FIG. 9A, there are four phrases, and the utterance times are Ps1, Ps2, Ps3, and Ps4.

The delimiter position / pause time determination unit 35 determines whether or not the delay time is less than or equal to the first phrase. In FIG. 9B, the first delimiter position is determined.
(i) It is determined whether Ps1 is equal to or less than Δt. When the determination is N, the position after “XXX” is a separation position. If the utterance time of one phrase is not less than the delay time, it may be divided into character units shorter than the phrase.

In addition, when the overlap time of the direct sound and the reflected sound is short, the intelligibility of the voice does not decrease so much, so that the delay overlap transmission for a time that does not affect the intelligibility of the voice is acceptable. In this case, if a time that does not affect the intelligibility of speech is α, Δt may be replaced with “Δt−α” in the following determination formula.
(ii) If the determination is Y, it is determined whether Ps1 + Ps2 is equal to or less than Δt. When this determination is N, the break position is after “XXX”.
(iii) If the determination is Y, it is determined whether Ps1 + Ps2 + Ps3 is equal to or less than Δt. When this determination is N, the position after “ΔΔΔΔ” is the separation position.
(iv) If the determination is Y, it is determined whether Ps1 + Ps2 + Ps3 + Ps4 is equal to or less than Δt. When this determination is N, the position after “□□” is a separation position.
When Ps1 + Ps2 + Ps3 + Ps4 is equal to or less than Δt, it is determined that no separation is necessary.

In FIG. 9C, when the delimiter position is after “XXX”, the second delimiter position is determined.
(i) It is determined whether Ps2 is equal to or less than Δt. When the determination is N, the position after “ΔΔΔΔ” is the separation position.
(ii) If the determination is Y, it is determined whether Ps2 + Ps3 is equal to or less than Δt. When this determination is N, the position after “ΔΔΔΔ” is the separation position.
(iii) If the determination is Y, it is determined whether Ps2 + Ps3 + Ps4 is equal to or less than Δt. When this determination is N, the position after “□□” is a separation position.
When Ps2 + Ps3 + Ps4 is equal to or less than Δt, it is determined that the second delimiter is unnecessary.

In FIG. 9D, when the delimiter position is after “ΔΔΔΔ”, the second or third delimiter position is determined.
(i) It is determined whether Ps3 is equal to or less than Δt. When the determination is N, the position after “□□” is a separation position.
(ii) If the determination is Y, it is determined whether Ps3 + Ps4 is equal to or less than Δt. When this determination is N, the position after “□□” is a separation position.
When Ps3 + Ps4 is equal to or less than Δt, it is determined that the second or third delimiter is unnecessary.

  As described above, the separation position can be determined for each maximum number of phrases whose total utterance time is shorter than the delay time, and the separation position can be determined for each phrase at the minimum.

  For example, the sentence length Ts of “Please evacuate to the hill” is “937.5 + 937.5 + 750 = 2625”. Since the utterance time of the first phrase (on the hill) is 937.5 [ms] and the delay time of the slave station with the slave station ID = 1 is 1000 [ms], it is summed with the utterance time of the next phrase (evacuated).

  Since “937.5 + 937.5 = 1875” is larger than the delay time, the break position is after the first phrase (on the hill).

  Next, since the utterance time of the second phrase (evacuation) is 937.5 [ms], it is totaled with the utterance time of the third phrase (please).

  Since “937.5 + 750 = 1687.5” is greater than the delay time, the break position is after the second phrase (evacuation). Therefore, in this voice message, a pause is inserted for each phrase.

The delimiter position / pause time determination unit 35 determines the pause time.
FIG. 10 is an example of a diagram illustrating the pause time. Since the utterance time of one phrase is secured during the pause time, the pause time is longer than the utterance time of the immediately preceding phrase. If the delay time and the phrase time (reflected sound) are not greater than or equal to the sum, a delay overlap transmission with the next phrase occurs. Accordingly, as shown in the figure, the pause time is a value obtained by adding the phrase utterance time to the difference between the delay time Δt and the phrase utterance time Ps. This value is equal to the delay time Δt.

  The pause time is set to Δt−α when the overlap of the time α that does not affect the intelligibility of the speech is allowed. By shortening the pause time, it is possible to reduce the time until transmission of the entire voice message is completed.

  Although it is possible to secure the pause time longer than the delay time, if the pause time is unnecessarily long, it takes time to complete the transmission of the voice message, which is not preferable on the disaster prevention radio system. However, if the urgency is low, a pause time longer than the delay time may be secured.

  In addition, some delay overlap transmission (for example, about 1 mora) does not reduce the intelligibility so much, so it does not exclude that the pause time is less than the delay time.

  The break position / pause time determination unit 35 sends the break position and pause time to the pause insertion unit 36. The break position / pause time determination unit 35 may reconstruct the voice message by inserting a pause of the determined pause time at the break position. In this embodiment, it is assumed that the pose insertion unit 36 inserts a pose into the text data divided by the phrase division unit 33.

  The pause insertion unit 36 inserts a pause time comparable to the delay time Δt into the original voice message. Specifically, a blank character (space) is inserted. Since the utterance time of one blank character is determined, the number of blank characters corresponding to the pause time is inserted. As described above, a voice message in which a pose is inserted such as “on a hill”, “pause”, “evacuate”, “pause”, and “please” is obtained.

  The pose insertion device 30 transmits a voice message in which a pose is inserted to the console 20. The console 20 converts the voice message in which the pause is inserted into a voice signal again. For example, it may be displayed on the display unit or printed, and the user speaks again with a pause and collects sound from the microphone 11. Alternatively, the speech signal may be acquired from the speech synthesizer 23 by being transmitted to the speech synthesizer 23.

  Since the transmission unit 17 transmits a voice message in which a pause of about the delay time is inserted, it is possible to suppress the delay and overlap transmission of the voice that the slave station 200 utters from the speaker 53.

[Modification]
In FIG. 7, a pause is inserted into a voice message converted into a voice signal, but the timing for inserting a pause may be set to an appropriate timing in practice.

  FIG. 11 shows another example of a functional block diagram of the pose insertion device 30. FIG. 11A is a functional block diagram when a voice message is given as text data (electronic data or electronic file). In FIG. 11A, text data is given instead of the voice signal uttered by the user, so that there is a voice message acquisition unit 41 instead of the voice signal acquisition unit 31. The voice message acquisition unit 41 acquires a voice message suitable for emergency information from the console 20 or the network 24. When acquiring from the console 20, there are text data of a voice message selected by the user, text data automatically selected according to a notification from the telemeter 300, text data instructed by J-Alert, and the like.

  If the voice message is text data, voice recognition is not necessary, and the voice recognition unit 32 is not necessary. That is, the text data of the voice message acquired by the voice message acquisition unit 41 is directly sent to the phrase dividing unit 33. The subsequent procedure is the same as in FIG.

  FIG. 11B is a functional block diagram when a voice message is given as being printed on paper. The content of the voice message may be transmitted by FAX or handed to the user once it has been launched. In this case, it may be converted into a voice signal by the user uttering as shown in FIG. 7, but there is a need for voice recognition or the like, and there is a possibility that a recognition error may be uttered. For this reason, when a voice message is given as being printed on paper, it may be effective to perform the following processing.

  In FIG. 11B, since a voice message is given as a document, a scanner device 42 that optically reads the document is used. The scanner device 42 may be connected to the pose insertion device 30 or may be connected to the console 20. In any case, the pose insertion device 30 can acquire image data of a document. The scanner device 42 generally has a line sensor in which image sensors are arranged one-dimensionally. However, a document may be photographed with a so-called camera (two-dimensional image sensor) to create image data.

  The pose insertion device 30 has an OCR (Optical Character Reader) unit 43, and converts a voice message described in image data into text data. Therefore, text data is obtained in the same manner as in FIG. The subsequent processing is the same as in FIG.

  In addition, it is also possible to perform the same processing after translating a voice message (voice signal, text data, manuscript) into a foreign language. When many foreigners live in the area where the slave station 200 is laid, the voice message can be expanded in the native language of the foreigner.

  As described above, there is no limitation on how the voice message is input to the pause insertion device 30. If the input voice message is converted into text data, the processing of the present embodiment can be similarly performed. .

[Operation procedure]
FIG. 12A is an example of a flowchart illustrating an operation procedure of the pose insertion device 30. First, the pause insertion device 30 acquires a voice message conveying the contents of emergency information in the form of text data (S10). That is, the voice recognition unit 32 recognizes the voice signal acquired by the voice signal acquisition unit 31, the voice message acquisition unit 41 receives the text data itself, or the scanner scans the document and the OCR unit 43 sets the text data. Obtained by converting to.

  Next, the phrase dividing unit 33 divides the text data into phrases (S20).

  The utterance time length calculation unit calculates the utterance time of each phrase (S30).

  The delimiter position / pause time determination unit 35 determines the delimiter position of the text data (S40). As described above, one or more phrases whose total utterance time Ps is equal to or less than the delay time Δt are collected to determine a break position.

  The delimiter position / pause time determination unit 35 determines the pause time (S50). The pause time is about the same as the delay time Δt.

  The pause insertion unit 36 inserts an unvoiced character (space) of the same degree as the delay time at the break position determined by the break position / pause time determination unit 35 (S60).

<Modification of operation>
As described so far, instead of inserting a pause (unvoiced character or silent time) into the voice message, the delimiter position / pause time determination unit designates only the pause position and pause time, and the control unit 15 selects the slave station 200. A pause may be inserted when sending a voice message to.

  FIG. 12B is an example of a flowchart illustrating an operation procedure of the pose insertion device 30. The processing up to step S50 is the same, but at S55, the delimiter position / pause time determination unit attaches the delimiter position and pause time to the converted message (S55). For example, with respect to the text length of the converted message, a delimiter position is specified at time t with reference to the time when utterance is started, and a pause time for each delimiter position is designated. For example, when there are two pause positions, “(t1, Δt), (t2, Δt)” is obtained.

  The console receives a voice message, a break position and a pause time from the pause insertion device (S110).

  The control unit of the console starts transmitting a voice message (S120).

  While transmitting the voice message, the control unit determines whether or not the time at the break position has elapsed from the start of the voice message utterance (S130). Until the time of the delimiter position is reached (No in S130), the control unit determines whether or not the transmission of the voice message is finished (S170), and if not finished, the transmission of the voice message is continued.

  When the time at the delimiter position is reached (Yes in S130), the control unit stops the transmission of the voice message (S140). Then, it is determined whether or not the pause time has elapsed until the pause time has elapsed (S150).

  When the pause time has elapsed (Yes in S150), the control unit resumes transmitting the voice message from the interrupted location (S160). This part is a phrase and a phrase separation position. The control unit is not aware of the unit of phrase, but as a result, a pause is inserted between phrases. In such a transmission method, it is not necessary to insert a pause (unvoiced character or silent time) in the voice message, and the pause can be inserted in the transmission process.

  As described above, the disaster prevention radio system 100 according to the present embodiment divides a voice message and inserts a pause, so that it is possible to amplify a voice message that can be easily heard by local residents in an echo environment.

  In the first embodiment, the delimiter position and pause time are determined according to the delay time obtained in advance for each slave station. However, as shown in the above experimental results, the accuracy rate does not always decrease as the delay time increases. Although not in the experimental results, according to the applicant's research, it has become clear that the correct answer rate differs depending on the sentence length even if the delay time is the same. Moreover, since the transmission path of the reflected sound transmitted from the slave station to the area is not necessarily one, the delay time may not be specified uniquely.

  Therefore, in this embodiment, the delay time that most affects the intelligibility reduction rate is determined according to the relationship between the measured delay time and the intelligibility reduction rate, and the delimiter position and pause time are determined according to the delay time. The disaster prevention radio system 100 to be determined will be described. The intelligibility reduction rate indicates a rate (%) at which the correct answer rate decreases according to the delay time when the correct answer rate at which the subject correctly answers the phrase when the delay time is zero is 100.

  FIG. 13 shows an example of a functional block diagram of the pose insertion device 30. The pause insertion apparatus of FIG. 13 is configured such that the delay time database 38 of FIGS. 7 and 10 is replaced with a long path echo calculation unit 61, a delay time determination unit 62, and an echo / intelligibility correspondence database 63. The delay time determination unit 62 uses the information from the long path echo calculation unit 61 and the echo / intelligibility correspondence database 63 to determine the delay time that most affects the intelligibility reduction rate in each slave station.

  FIG. 14 (a) shows the state of voices simultaneously emitted from three slave stations at a certain point. The horizontal axis is time, and the vertical axis is relative amplitude (Relative Amplitude). The control unit 52 of the slave station 200 uses the reference to the time when the sound is amplified from the slave station 200 (time t0), and the sound that the microphone 54 emits from another slave station at time t1 (the echo intensity of the claims) Measure delay time Δt until it is detected as an example. Time t2 is the time when the voice from another slave station was measured. In this figure, Δt is determined using sound from other slave stations, but there are obstacles such as mountains and large buildings in the environment that are not affected by other slave stations. In some cases, Δt is determined by the reflected sound from each obstacle. The slave station transmits the signal waveform itself of FIG. 14A or the delay times Δt1 and Δt2 to the disaster prevention radio system 100.

  When a plurality of delay times are measured, it is conceivable to adopt a delay time having the largest relative amplitude. However, even if a pause is inserted based on the delay time with the largest relative amplitude, there is a possibility that the improvement of the intelligibility is not necessarily affected (the effect is low). This is because the degree of intelligibility is also affected by human sensitivity characteristics and sentence length.

  FIG. 14B is an example of a diagram for schematically explaining an example of the echo / intelligibility correspondence database 63. For each sentence length, an arbitrary delay time is associated with an intelligibility reduction rate. That is, in the figure, the intelligibility reduction rate is associated with delay times every 300 [ms], but the relationship between the delay time and the intelligibility reduction rate is registered at intervals that are short enough to be interpolated. Therefore, the delay time determination unit 62 can determine the intelligibility reduction rate for any delay time.

  Focusing on one sentence length, the intelligibility reduction rate increases up to a certain range of delay time, but the intelligibility reduction rate may decrease as the delay time further increases. For example, at a sentence length of 2000 [ms], the intelligibility reduction rate increases until a delay time of 900 [ms], but decreases at 1200 [ms]. Thus, the intelligibility reduction rate does not necessarily decrease as the delay time Δt increases.

  When attention is paid to the same delay time, the intelligibility reduction rate differs depending on the sentence length. For example, when the delay time is 600 [ms], the intelligibility reduction rate is 10% at a sentence length of 1000 [ms], the intelligibility reduction rate is 13% at a sentence length of 2000 [ms], and the sentence length is 3000 [ ms], the intelligibility reduction rate is 16%.

  Therefore, when the long path echo calculation unit 61 calculates a plurality of delay times, it can be said that it is preferable to determine the delay time for determining the break position where the pause is inserted based on the sentence length and the intelligibility reduction rate. . For example, when the delay time of 300 [ms] and 600 [ms] is measured when the sentence length is 1000 [ms], the delay time calculation unit 42 delays 600 [ms] having the highest intelligibility reduction rate. The time Δt is determined. When the sentence length is 2000 [ms] and the delay times of 300 [ms], 600 [ms], and 900 [ms] are measured, the delay time calculation unit 42 has the highest intelligibility reduction rate 900 [ms]. ] Is determined as the delay time Δt. By doing so, it is possible to determine the delay time that most affects the intelligibility reduction rate among the measured delay times.

  FIG. 15 is an example of a flowchart showing an operation procedure of the pose insertion device 30. In FIG. 15, the delay time is calculated in step S35. Processing other than this step is the same as in FIG.

  As described above, the disaster prevention radio system according to the present embodiment prevents the voice message from being overlapped and transmitted to the area where the voice message is transmitted through a plurality of transmission paths with the delay time having the highest intelligibility reduction rate. , You can insert a pose.

20 console 21 master station communication device 24 network 30 voice conversion device 33 phrase division unit 34 utterance time length calculation unit 35 delimiter position / pause time determination unit 36 pause insertion unit 53 speaker 100 disaster prevention radio system 200 slave station

Claims (6)

An outdoor environmental voice transmission device that transmits voice messages for disaster prevention to one or more slave stations laid in the area,
Voice message acquisition means for acquiring the voice message;
Message dividing means for dividing the voice message acquired by the voice message acquiring means into a meaningful symbol string or requesting the outside to divide the voice message into the symbol string;
Utterance time calculating means for calculating the utterance time of each symbol string;
A delay time database in which the delay time of echo generated when the voice message is amplified from a slave station is registered for each slave station;
For each of the one or more symbol strings whose total utterance time is equal to or less than the delay time, a process for determining the end of the symbol string as the break position of the voice message is transmitted to the slave station registered in the delay time database Delimiter position determining means for performing a voice message based on the delay time of the slave station;
Unvoiced character insertion means for inserting unvoiced characters whose utterance time is longer than the delay time at the break position of the voice message;
Message transmission means for transmitting a voice message with unvoiced characters inserted to a slave station;
An outdoor environment audio transmission device comprising: An outdoor environmental voice transmission device that transmits voice messages for disaster prevention to one or more slave stations laid in the area,
Voice message acquisition means for acquiring the voice message;
Message dividing means for dividing the voice message acquired by the voice message acquiring means into a meaningful symbol string or requesting the outside to divide the voice message into the symbol string;
Utterance time calculating means for calculating the utterance time of each symbol string;
A delay time database in which the delay time of echo generated when the voice message is amplified from a slave station is registered for each slave station;
Separation position for determining the end of one or more of the symbol strings whose total utterance time is equal to or less than the delay time for each slave station registered at least in the delay time database as the separation position of the voice message A determination means;
Message transmitting means for transmitting the voice message and stopping the transmission of the voice message when the delimiter position of the voice message is detected, and restarting the transmission of the voice message when a delay time elapses after the stop. ,
An outdoor environment audio transmission device comprising: An intelligibility database in which intelligibility information for correctly understanding the meaning of the symbol string for different delay times is registered;
A delay time determining means for determining a delay time at which the intelligibility information is reduced most in the intelligibility database based on one or more delay time information transmitted from each slave station, as a delay time of the slave station; Have
The delimiter position determining means determines the delimiter position of the voice message to be transmitted to each slave station based on the delay time determined by the delay time determining means.
The outdoor environment audio transmission apparatus according to claim 1 or 2, In the intelligibility database, for each utterance time of the entire voice message, intelligibility information for correctly understanding the meaning of the symbol string for different delay times is registered,
The delay time determining means determines a delay time of a slave station based on the delay time associated with the utterance time of the entire voice message to be transmitted and the intelligibility information.
The outdoor environment audio transmission device according to claim 3. The unvoiced character insertion means makes the utterance time of unvoiced characters to be inserted at the delimiter position shorter than the delay time for a time that does not affect the intelligibility of the symbol string more than a predetermined amount,
The outdoor environment audio transmission apparatus according to claim 1. An outdoor environment audio transmission device that transmits a voice message for disaster prevention to one or more slave stations laid in the area, and an outdoor environment audio transmission system having one or more slave stations,
Voice message acquisition means for acquiring the voice message;
Message dividing means for dividing the voice message acquired by the voice message acquiring means into a meaningful symbol string or requesting the outside to divide the voice message into the symbol string;
Utterance time calculating means for calculating the utterance time of each symbol string;
A delay time database in which the delay time of echo generated when the voice message is amplified from a slave station is registered for each slave station;
For each of the one or more symbol strings whose total utterance time is equal to or less than the delay time, a process for determining the end of the symbol string as the break position of the voice message is transmitted to the slave station registered in the delay time database Delimiter position determining means for performing a voice message based on the delay time of the slave station;
Unvoiced character insertion means for inserting unvoiced characters whose utterance time is longer than the delay time at the break position of the voice message;
Message transmitting means for transmitting a voice message in which unvoiced characters are inserted to a slave station,
The slave station is
Message receiving means for receiving a voice message with unvoiced characters inserted;
A loudspeaker for loudening a voice message in which an unvoiced character is inserted,
An outdoor environment audio transmission system characterized by that.

Images