JP6036339B2 - Speech device - Google Patents

Description

  The present invention relates to an utterance device that utters a guidance voice or the like.

  DESCRIPTION OF RELATED ART Conventionally, the speech apparatus which utters a guidance voice automatically like a navigation apparatus, a parking assistance apparatus, or an automatic cash transaction apparatus is known (for example, patent document 1).

  In the utterance device, a vast number of recorded audio data is stored in a storage device such as an HDD (Hard disk drive) or an SD card (SD Memory Card). Then, the utterance device searches and reads out the audio data requested by the application program (hereinafter referred to as application) from the storage device, and utters the audio according to the application.

  However, the access time to the storage device takes longer than the CPU processing time. Therefore, if all the audio data stored in the storage device is searched one by one, the cumulative number of accesses increases. And since the number of times of cumulative access increases, it takes time from when the application requests voice data until it is actually uttered, resulting in poor response.

  Therefore, in order to solve this problem, the following has been conventionally performed. First, secondary index data classified for a predetermined item such as language, which is an index of audio data, and primary index data classified for a predetermined item such as language, which is an index of secondary index data, Is stored in the storage device. Subsequently, only the primary index data of the primary index data and the secondary index data is expanded in the cache memory. This is because the capacity of the cache memory is smaller than that of the storage device, and the amount of data to be expanded must be suppressed.

  When the necessary audio data is read from the storage device, the primary index data corresponding to the necessary audio data is searched from the primary index data expanded in the cache memory. Furthermore, referring to the retrieved primary index data, secondary index data corresponding to necessary audio data is retrieved from the secondary index data stored in the storage device. Then, referring to the retrieved secondary index data, necessary audio data is accessed and read. This reduces the chances of searching the data stored in the storage device and reduces the number of accesses to the storage device, thereby speeding up the time from when the app requests voice data until it actually speaks. I was trying to make it.

  In addition, the hierarchical data structure such as audio data, primary index data, and secondary index data is used to reduce the overall data size and to reduce the trouble of editing when adding new audio data. It is.

JP 2006-311128 A

  However, in the above-described technique in which only the primary index data of the primary index data and the secondary index data is expanded in the cache memory, the storage device is searched for the secondary index data stored in the storage device. There was a problem that access time would always occur.

  In addition, when secondary index data is to be expanded in the cache memory, it is necessary to greatly increase the cache memory capacity. However, if the cache memory is increased, the access time of the cache memory increases. Will occur.

  The present invention has been made in view of the above-described conventional problems, and its object is to reduce the time from when an application program requests voice data until it actually speaks while suppressing an increase in the capacity of the cache memory. An object of the present invention is to provide an utterance device that can further increase the speed.

  The utterance device of the present invention includes a plurality of audio data, secondary index data classified by language as an index of audio data, and primary index data classified by language as an index of secondary index data. The storage (30) that is stored, the cache memory (14) having an access latency smaller than that of the storage, the reading unit (113) that reads target audio data that is the audio data requested by the application program from the storage, and the reading unit An utterance device including an utterance unit (117) that utters by outputting a sound according to the read target audio data, and caches only primary index data of a language in which language is set among primary index data Secondary index data in the same language as the primary index data expanded in memory and expanded in cache memory Among them, a cache unit (111) that narrows down to a predetermined number of secondary index data and expands the cache memory, and target secondary index data that is secondary index data that is an index of target audio data requested by the application program, A cache memory search unit (112) for searching from secondary index data expanded in the cache memory, and the reading unit searches the secondary index data expanded in the cache memory for target secondary index data by the cache memory search unit If it can, the target audio data is read from the storage with reference to the target secondary index data expanded in the cache memory.

  According to this, since it is expanded to the cache memory by focusing only on the primary index data of the language for which the language is set (hereinafter referred to as “set language”), the secondary index data of the set language can be stored without greatly increasing the cache memory. Part of it can be expanded in cache memory. If the language setting is not switched, primary index data other than the set language (hereinafter, non-set language) is not required for reading the audio data. Therefore, even if the primary index data of the set language is narrowed down and expanded in the cache memory, it is not necessary to access the primary index data of the non-set language stored in the storage, and the primary index data stored in the storage No access processing time is required.

  In addition, since a predetermined number of data among the secondary index data in the set language is expanded in the cache memory, the target secondary index data serving as an index of the target audio data is included in the secondary index data expanded in the cache memory. In this case, there is no need to search for secondary index data stored in the storage. In this case, it is not necessary to access the secondary index data stored in the storage, and the access processing time for the secondary index data stored in the storage can be avoided.

  Note that the predetermined number here refers to the predetermined number of secondary index data in the set language relative to the memory capacity required when all the primary index data in the non-set language is expanded in the cache memory. The number of memory increases required when deploying to a number that falls within the range of several kilobytes.

  Since the cache memory has an access latency smaller than that of the storage, and the access time to the cache memory is shorter than the access time to the storage, the above configuration increases the opportunity to reduce the access processing time. As a result, it is possible to further speed up the time from when the application program requests voice data until it actually utters while suppressing an increase in the capacity of the cache memory.

1 is a diagram illustrating an example of a schematic configuration of a navigation device 1. FIG. It is a schematic diagram which shows an example of the data structure of an audio | voice data file. 2 is a functional block diagram illustrating an example of a schematic configuration of a CPU 11. FIG. 3 is a schematic diagram for explaining a memory area of a RAM 14; FIG. It is a flowchart which shows an example of the flow of the cache process in CPU11. It is a flowchart which shows an example of the flow of the read related process in CPU11.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. In this embodiment (hereinafter referred to as Embodiment 1), an example in which the utterance device of the present invention is applied to the navigation device 1 will be described. A navigation device 1 shown in FIG. 1 is mounted on a vehicle and includes an audio output device 20, a storage device 30, and a control device 10 connected thereto. Here, for the sake of convenience, among the configurations related to the functions of a general navigation apparatus, those not necessary for the description of the present invention are not described.

  The voice output device 20 is composed of a speaker or the like, and utters a guidance voice or the like at the time of route guidance according to the voice data output from the control device 10.

  The storage device 30 is an electrically writable storage device. The storage device 30 has uses such as storing a large amount of data and data that should not be erased even when the power is turned off. The storage device 30 may be a removable memory having a relatively small storage capacity. For example, the storage device 30 may be configured to use an HDD (Hard disk drive), an SD card (SD Memory Card), or the like.

  Further, the storage device 30 stores voice data files (that is, archives) that are guidance voice data uttered from the voice output device 20 and include a plurality of actual voice data (hereinafter simply referred to as voice data). The storage device 30 corresponds to the storage in the claims. For example, the audio data is compressed in a predetermined format and stored in the storage device 30.

  Here, the data structure of the audio data file stored in the storage device 30 will be described with reference to FIG. The audio data file has a hierarchical data structure such as audio data, an audio offset table as an index of audio data, and an audio management record as an index of the audio offset table. The arrangement of the voice management table A in FIG. 2 shows the data of the hierarchy of the voice management record, and the arrangement of the voice management table B in FIG. 2 shows the data of the hierarchy of the voice offset table.

  The number of speech languages indicates the number of languages used for speech data. As an example, languages used for audio data include Japanese, English, French, Spanish, Dutch, Italian, German, Chinese, Korean, and the like.

  A voice management record is provided for each language. In other words, it is classified by language. For example, when the number of languages is n, voice management records 1 to n are included in the voice data file. The voice management record provided for each language corresponds to the primary index data in the claims.

  The voice management record includes items such as the number of voices, an offset to the voice offset table, and a size of the voice offset table. The offset to the audio offset table and the size of the audio offset table are values indicating one specific offset table among the plurality of offset tables. For example, one record of the voice management record is about several tens of bytes.

  An audio offset table is also provided for each language. In other words, it is classified by language. For example, when the number of languages is n, the audio data table includes 1 to n audio offset tables.

  In addition, each audio offset table for each language includes an audio offset record. The audio offset record is provided for each ID (hereinafter referred to as audio ID) that specifies audio data. For example, when there are l pieces of audio data for one language, 1 to 1 audio offset records are included in one audio offset table. Note that a voice offset record classified for each language and provided for each voice ID corresponds to the secondary index data in the claims.

  The audio offset record includes an audio ID item, an offset to audio data, and an audio data size item. The offset to the audio data and the size of the audio data are values indicating one specific audio data among the plurality of audio data. For example, one record of the audio offset record is about 10 to 20 bytes. Further, the number of audio offset records is, for example, about 7500 in one audio offset table.

  Furthermore, the audio offset record is stored so that the audio offset record that is an index of audio data having a higher use frequency is closer to the head in the storage order in each audio offset table.

  For example, when there are k pieces of sound data, the sound data of 1 to k is stored in the sound data file. The audio data is, for example, about several tens to several hundreds of kilobytes per record.

  Note that the audio data file has a hierarchical data structure such as an audio management record, audio offset table, and audio data in order to keep the overall data size small and to edit when adding new audio data. This is to reduce the time and effort. Specifically, when new audio data is added, the audio management record is not changed, and a new audio offset record for indexing the added audio data only needs to be added to the audio offset table. .

  The control device 10 is configured as a normal microcomputer, and includes a well-known CPU 11, a memory 12 such as a ROM 13 and a RAM 14, an I / O (not shown), and a bus line for connecting these configurations. As illustrated in FIG. 3, the CPU 11 includes a cache unit 111, a cache memory search unit 112, a read unit 113, a key acquisition unit 114, a storage access unit 115, a storage search unit 116, and a voice generation unit 117 as functional blocks. Yes.

  The CPU 11 executes a cache-related process for caching a part of the voice data file in the RAM 14 and a read-related process related to a process for reading voice data requested by the voice guidance application described later from the storage device 30. Note that cache processing and read-related processing will be described in detail later.

  The ROM 13 is a nonvolatile memory and stores a program to be executed by the CPU 11. For example, the ROM 13 stores an application program (hereinafter referred to as a voice guidance application) that utters a guidance voice at a predetermined timing, an application that manages language settings (hereinafter referred to as a language management application), and the like.

  The language setting indicates setting a language used for voice guidance from a plurality of languages. Information about the language set language (hereinafter, currently set language) managed by the language management application and the sequence number of each language in the audio data file (hereinafter, language sequence number) is, for example, RAM 14 or Assume that it is stored in an EEPROM (not shown) of the memory 12.

  The RAM 14 is a volatile memory having an access latency smaller than that of the storage device 30. For example, a DRAM (Dynamic Random Access Memory) can be used as the RAM 14. Here, the memory area of the RAM 14 will be described with reference to FIG. As shown in FIG. 4, the RAM 14 has a program area and a RAM area as memory areas. The program area is an area where a program stored in the ROM 13 is copied. The RAM area is an area other than the program area in the memory area of the RAM 14.

  As shown in FIG. 4, the above-described voice guidance application is copied to the program area. In addition, as shown in FIG. 4, a memory area for use of the voice guidance application is allocated to the RAM area, and a part of the voice management record and a part of the voice offset table are cached in the memory area. . Therefore, the RAM 14 corresponds to the cache memory of the claims. In addition, the audio data decoded by the CPU 11 is temporarily stored in the RAM area.

  Here, the cache processing in the CPU 11 will be described using the flowchart of FIG. The flowchart of FIG. 5 may be configured to start when the voice guidance application is activated or when the language setting is switched.

  In addition, when the navigation device 1 is activated in conjunction with the ignition power supply or accessory power supply of the vehicle on which the navigation device 1 is mounted and the voice guidance application is automatically activated, the ignition device is automatically activated. The configuration shown in FIG. 5 may be started when the power source or the accessory power source is turned on. In addition, the language setting may be switched by the CPU 11 that executes the language management application.

  First, in step S1, the cache unit 111 performs language setting related information acquisition processing, and proceeds to step S2. In the language setting related information acquisition process, the cache unit 111 acquires the current setting language and language sequence number managed by the language management application. The current set language and language sequence number may be obtained from the above-described RAM 14 or EEPROM.

  In step S2, the cache unit 111 performs the first cache process, and proceeds to step S3. In the first cache process, an audio data file stored in the storage device 30 is accessed. Then, based on the current set language and language sequence number acquired in step S1, only the voice management record for the current set language among the voice management records stored in the storage device 30 is expanded in the RAM. That is, only the voice management record for the current set language is cached in the RAM 14.

  In step S3, the cache unit 111 performs a first key acquisition process, and proceeds to step S4. In the first key acquisition process, the voice management record cached in the RAM 14 in step S2 (hereinafter, cached voice management record) is accessed. Then, an offset to the audio offset table (hereinafter referred to as table offset) and the size of the audio offset table (hereinafter referred to as table size) included in the cache audio management record are acquired.

  In step S4, the cache unit 111 performs the second cache process and ends the flow. In the second cache process, an audio data file stored in the storage device 30 is accessed. Then, based on the table offset and table size acquired in step S3, only a predetermined number of audio offset records in the audio offset table for the current setting language stored in the storage device 30 are expanded in the RAM. That is, only a predetermined number of audio offset records for the currently set language are cached in the RAM 14.

  The predetermined number here may be set in advance so as to satisfy the following conditions. Increase in the memory capacity required when the audio offset record is cached with respect to the difference in memory capacity when the cache audio management record is cached in the RAM 14 as compared to when the entire audio management record of the audio data file is cached in the RAM 14. The number should be enough to fit a few kilobytes. As an example, the predetermined number may be several thousand.

  In addition, as a method of selecting a predetermined number of audio offset records, a configuration in which a predetermined number of audio offset records that serve as an index of a predetermined number of audio data having a higher use frequency may be selected. Specifically, a predetermined number of audio offset records may be selected in order from the storage order in the audio offset table close to the head. This is because, as described above, in the audio offset table, the audio offset records that are indexes of audio data having a higher use frequency are stored such that the storage order is closer to the head. Note that a predetermined number may be selected at random.

  Subsequently, the read-related processing in the CPU 11 will be described using the flowchart of FIG. The flowchart in FIG. 6 may be configured to start when the above-described cache processing is completed.

  First, in step S11, when there is a request for voice data from the voice guidance application (hereinafter referred to as an utterance request) (YES in step S11), the process proceeds to step S12. On the other hand, if there is no utterance request (NO in step S11), the process proceeds to step S20. The utterance request includes the voice ID of the requested voice data as an argument.

  In step S12, the cache memory search unit 112 performs a cache offset record search process, and proceeds to step S13. In the cache offset record search process, a voice offset record including a voice ID that matches the voice ID in the utterance request argument from the voice offset record cached in the RAM 14 by the cache process based on the voice ID in the utterance request argument. Explore.

  In the cache offset record search process, when the search target voice offset record can be searched, the offset from the voice offset record to the voice data requested to be uttered (hereinafter, data offset), and the utterance request Acquire the size of the audio data that has been recorded (hereinafter referred to as the data size).

  In step S13, when the data offset and the data size can be acquired in the cache offset record search process (YES in step S13), the process proceeds to step S14. On the other hand, when the data offset and the data size cannot be acquired (NO in step S13), the process proceeds to step S15.

  In step S14, the reading unit 113 performs a first audio data reading process, and proceeds to step S19. In the first audio data reading process, an audio data file stored in the storage device 30 is accessed. Then, the audio data indicated by the data offset and the data size acquired by the cache offset record search process is read from the storage device 30. In other words, the audio data is read from the storage device 30 with reference to the data offset and data size acquired in the cache offset record search process.

  In step S15 when the data offset and data size cannot be acquired in step S13, the key acquisition unit 114 performs the second key acquisition process, and proceeds to step S16. In the second key acquisition process, the cache voice management record is accessed. Then, the table offset and table size included in the cache voice management record are acquired.

  In step S16, the storage access unit 115 performs storage access processing, and proceeds to step S17. In the storage access process, an audio data file stored in the storage device 30 is accessed. Then, the audio offset table indicated by the table offset and the table size acquired in the second key acquisition process is accessed.

  In step S17, the storage search unit 116 performs storage search processing, and proceeds to step S118. In the storage search process, a voice offset record including a voice ID that matches the voice ID in the argument of the utterance request is searched from the voice offset table accessed in the storage access process based on the voice ID in the argument of the utterance request. . In the storage search process, when the search target audio offset record can be searched, the data offset and the data size are acquired from the audio offset record.

  In step S18, the reading unit 113 performs the second audio data reading process, and proceeds to step S19. In the second audio data reading process, an audio data file stored in the storage device 30 is accessed. Then, the audio data indicated by the data offset and data size acquired in the storage search process is read from the storage device 30. In other words, the audio data is read from the storage device 30 with reference to the data offset and data size acquired in the storage search process.

  In step S19, the utterance unit 117 performs utterance processing, and the process proceeds to step S20. In the utterance process, the reading unit 113 decodes the audio data read from the storage device 30 and outputs the decoded data to the audio output device 20, and outputs the audio from the audio output device 20. That is, the voice output device 20 is uttered.

  In step S20, if it is the end timing of the read-related process (YES in step S20), the flow ends. On the other hand, if it is not the end timing of the reading related process (NO in step S20), the process returns to step S11 and the flow is repeated.

  An example of the end timing is when the language setting is switched or when the voice guidance application is ended. In addition, when the ignition device or the accessory power source of the vehicle in which the navigation device 1 is mounted is turned off, the navigation device 1 is terminated in conjunction with the voice guidance application automatically. It is good also as a structure which makes an end timing the time when a power supply or an accessory power supply is turned off.

  Here, the effect in Embodiment 1 is demonstrated. In order to explain the operational effects compared to the prior art, the audio data reading process in the prior art will be explained. In the prior art, the data structure of the audio data file itself is assumed to be the same.

  In the prior art, when the voice guidance application is activated, the CPU 11 caches the voice management records for all languages of the voice data file in the RAM 14, but does not cache the voice offset records in the RAM 14. The utterance request includes the voice ID of the requested voice data and the language sequence number as arguments.

  Subsequently, when there is an utterance request, the CPU 11 determines, based on the language sequence number in the utterance request argument, the language that matches the language sequence number in the utterance request argument from the speech management record cached in the RAM 14. Search the voice management record of the line number. Then, the voice offset table offset and table size for the same language as the voice management record are acquired. For the subsequent processing, the same processing as in Steps S16 to S19 described above is performed.

  On the other hand, in the configuration of the first embodiment, a predetermined number of data in the audio offset record of the current set language is cached in the RAM 14. Therefore, when the audio offset record serving as an index of the audio data requested to be uttered is cached in the RAM 14, it is necessary to access and search the audio offset record of the audio file data stored in the storage device 30. Disappear. The RAM 14 has an access latency smaller than that of the storage device 30, and the access time to the RAM 14 is shorter than the access time to the storage device 30. Therefore, according to the configuration of the first embodiment, the opportunity to reduce the access processing time is increased as compared with the conventional technique.

  In the first embodiment, the voice management records in the currently set language are narrowed down to the RAM 14 and cached in the RAM 14, and the number of voice offset records in the currently set language is such that the increase in memory capacity is within the range of several kilobytes. Is cached in the RAM 14. Therefore, a part of the voice offset record of the current set language can be cached in the RAM 14 without significantly increasing the RAM 14.

  If the language setting is not switched, a voice management record in a language other than the set language (hereinafter, non-set language) is not required for reading the voice data. Therefore, even when the voice management record of the set language is narrowed down and cached in the RAM 14, it is necessary to access the voice management record of the non-set language stored in the storage device 30 as compared with the prior art. It does not occur. Therefore, there is no possibility that the access processing time increases as compared with the prior art.

  As a result, according to the configuration of the first embodiment, it is possible to further speed up the time from when the voice guidance application is requested to the actual utterance while suppressing the increase in the capacity of the RAM 14.

  Furthermore, in the configuration of the first embodiment, a predetermined number of audio offset records serving as indexes of audio data having a higher use frequency are cached in the RAM 14. Therefore, when there is an utterance request, it is more likely that the data offset and the data size for the voice data for which the utterance request has been made can be acquired from the cached audio offset record. Accordingly, it is possible to increase the opportunity for further speeding up the time from when the voice guidance application is requested to the actual voice production.

  In the first embodiment, an example in which the speech device of the present invention is applied to a navigation device has been described. However, the present invention is not necessarily limited thereto. For example, it is good also as a structure which applies the speech apparatus of this invention to the apparatus which utters guidance voice automatically, such as a parking assistance apparatus, an automatic cash transaction apparatus, and a vending machine. However, as described above, the utterance device of the present invention can further increase the time from when the application requests audio data until it actually utters. More preferably, the present invention is applied to a device that utters a guidance voice used in a severe vehicle. For example, the present invention is preferably applied to a vehicle navigation device that provides route guidance, a parking assistance device that provides parking assistance, and the like.

  The present invention is not limited to the above-described embodiments, and various modifications can be made within the scope of the claims, and the technical means disclosed in different embodiments can be appropriately combined. Such embodiments are also included in the technical scope of the present invention.

1 navigation device (voice device), 14 RAM (cache memory), 30 storage device (storage), 111 cache unit, 112 cache memory search unit, 113 reading unit, 117 voice unit

Claims (7)

Storage that stores a plurality of audio data, secondary index data classified by language, which is an index of the audio data, and primary index data classified by language, which is an index of the secondary index data (30),
A cache memory (14) having a lower access latency than the storage;
A reading unit (113) that reads out the target audio data, which is the audio data requested by the application program, from the storage;
An utterance device comprising: an utterance unit (117) that utters by outputting sound according to the target audio data read by the reading unit;
Of the primary index data, the primary index data of a language that is set in a language is expanded in the cache memory and the secondary index data in the same language as the primary index data expanded in the cache memory A cache unit (111) that expands the cache memory to a predetermined number of the secondary index data;
A cache memory search unit (112) for searching for the target secondary index data, which is the secondary index data serving as the index of the target audio data requested by the application program, from the secondary index data expanded in the cache memory; With
When the target secondary index data can be searched by the cache memory search unit from the secondary index data expanded in the cache memory, the reading unit expands the target secondary index data expanded in the cache memory. And reading out the target audio data from the storage. In claim 1,
The cache unit includes a predetermined number of the secondary indexes that serve as an index of a predetermined number of the audio data having a higher usage frequency among the secondary index data in the same language as the primary index data expanded in the cache memory. An utterance device characterized in that data is expanded in the cache memory. In claim 2,
The secondary index data stored in the storage is stored such that the secondary index data serving as an index of the audio data having a higher use frequency is closer to the top in the storage order,
The cache unit is a predetermined number that is an index of a predetermined number of the audio data whose storage order in the storage is closer to the top of the secondary index data in the same language as the primary index data expanded in the cache memory A speech device characterized in that the secondary index data is expanded in the cache memory. In any one of Claims 1-3,
The cache unit develops the primary index data of the language set in the primary index data in the cache memory when the application program is started, and develops the primary index in the cache memory. An utterance device, wherein the secondary index data in the same language as the data is expanded in the cache memory by focusing on a predetermined number of the secondary index data. In any one of Claims 1-4,
When the language setting is switched, the cache unit expands the primary index data of the language set in the primary index data to the cache memory, and expands the primary index expanded in the cache memory. An utterance device, wherein the secondary index data in the same language as the data is expanded in the cache memory by focusing on a predetermined number of the secondary index data. In any one of Claims 1-5,
When the target secondary index data cannot be searched by the cache memory search unit from the secondary index data expanded in the cache memory, the target secondary index data is stored in the storage. A storage search unit (116) for searching from the index data;
The reading unit, when the target secondary index data cannot be searched by the cache memory search unit from the secondary index data expanded in the cache memory, the target secondary index searched by the storage search unit An utterance apparatus that reads out the target audio data from the storage with reference to data. In any one of Claims 1-6,
An utterance device that is used in a vehicle and utters a guidance voice.

Images