JPH0376478B2 - - Google Patents

Description

[Detailed description of the invention]

The present invention relates to a method for editing a large number of types of speech phrases stored in a memory into a speech signal of a sentence specified by an input signal. In general, the audio signal of a sentence is stored in memory by sampling audio at an ultra-audible frequency (for example, 10 KHz) and dividing the wave height of each sample into digital values of 100 or more levels. Announcements are made at stations and airports. An example of a device for making such an announcement is disclosed in Japanese Patent Application Laid-Open No. 119810/1983. This is a sentence structure in which various clauses that are the constituent elements of multiple sentences are stored in a voice memory, and the addresses of the voice memory where the first part of each clause that makes up the sentence to be announced are stored are arranged in the order of announcement. The phrase arrangement for the sentence is stored in the sentence storage unit. Note that if there are multiple sentences to be announced, multiple phrase arrays for sentence construction are stored in the sentence storage section. When announcing a certain sentence, the contents of the phrase array for sentence construction corresponding to the sentence to be announced are sequentially read out from the sentence storage section, and the phrases are read out from the voice memory accordingly. By the way, in the case of station announcements, for example, "A train will arrive at Platform 1 soon." or "The train bound for Tokyo will depart from Platform 1." In most cases, if you change a clause such as ``Tokyo,'' the rest of the phrase can be used in common. However, with the above device, when announcing, for example, "A train will arrive at Platform 1 soon" and "A train will arrive at Platform 2 soon," it will not be possible to announce "Soon" or "On the first track." , "The train",
A phrase arrangement for sentence construction that memorizes the starting address for each of “I will arrive,” and a sentence structure that memorizes the starting address for each of “soon,” “on the second platform,” “the train,” and “arriving.” Since the phrase arrangement was memorized, a large-capacity text storage unit was required. Particularly when there are a large number of sentences to be announced, the increase in the capacity of the sentence storage section becomes remarkable. However, he did not create and memorize many types of sentences, and he could not remember them, such as ``from line 1'', ``from line 2'', ``from line 3'', ``to line 1'', ``to line 2'', ``to line 3'', and ``from line 3''.
To Tokyo”, “To Osaka”, “To Nagoya”, “Soon”, “
The memory capacity can be reduced by storing phrases such as "The train is departing", "The train is leaving", and "The train is arriving" in the memory, and editing and announcing these phrases. The present invention aims to provide the above-mentioned editing method and to reduce memory capacity. Hereinafter, the present invention will be explained in detail based on one embodiment shown in the drawings. This example is ``soon.''
XXXX, the train is arriving. ” and the first sentence pattern is “YYYY, ZZZZ, the train is leaving.”
It announces the second sentence pattern of
In XXXX, the phrase “on the 1st line,” “on the 2nd line,” or “on the 3rd line” is inserted, and in YYYY, “1st line” is inserted.
"From Platform 2", "From Platform 2", or "From Platform 3"
The phrase ``bound for Tokyo,'' ``bound for Nagoya,'' or ``bound for Osaka'' shall be inserted in ZZZZ. Hereinafter, XXXX, YYYY, and ZZZZ will be referred to as variable clauses, and the other clauses will be referred to as fixed clauses. In Figure 1, 1 is the voice memory, and each of the above phrases is the first
They are stored in the relationships shown in the table.

【table】

[Table] Each of these phrases is sampled at, for example, 10KHz, each sample is expressed in the form of information with an appropriate number of bits, and the segments that make up each phrase have segment numbers that represent their memory addresses. It is attached. Reference numeral 2 denotes a phrase table which stores the starting segment number of each memory phrase and the number of constituent segments of the memory phrase in the format shown in Table 2. In this memory, the clause inserted into XXXX,
The clause inserted into YYYY and the clause inserted into ZZZZ are stored at the respective connected clause addresses. For example, "on line 1" inserted in XXXX,
The first segment and number of segments for "on the 2nd line" and "on the 3rd line" are stored in order from addresses 1 to 3,
The leading segments and number of segments of "from line 1", "from line 2", and "from line 3" to be inserted into YYYY are stored at addresses 11 to 13.
The same applies to the clauses inserted into ZZZZ.

[Table] 3 is a sentence structure table in which information on the clauses that make up the first sentence pattern and the second sentence pattern is stored in order at each sentence pattern address as shown in Table 3.

[Table] Regarding the first sentence pattern, it consists of four clauses, and the first clause is a fixed clause, ``soon,'' so the fixed/variable memory section at address 1 contains ``0'', which represents a fixed clause. Since there is a second clause following the first clause, "0" is stored in the end storage section, indicating that there is a subsequent clause.
The clause address storage section stores clause address 7 of the clause table 2, which corresponds to "soon". Second
Since the bunsetsu is a variable clause from "on the first line" to "on the third line", "1" representing a variable clause is stored in the fixed/variable storage section at address 2, and the third clause follows the second clause. Therefore, "0" is stored in the end storage section, and "on line 1", which is stored at the beginning of "on line 1" to "on line 3" in phrase table 2, is stored in the clause address storage section. Clause address 1 for the phrase is stored as a representative. Since the third phrase is a fixed phrase "train ga", "0" is stored in the fixed/variable memory section at address 3, and since there is a fourth phrase following the third phrase, the end memory section is stored as "0". "0" is stored, and "Train ga" is stored in the phrase address storage section.
Bunsetsu address 10 corresponding to is stored. The fourth clause is a fixed clause, “I will arrive.”
"0" is stored in the fixed/variable storage section at the 4th address, and since the fourth clause is the final clause, "1" representing the final clause is stored in the end storage section, and "1" representing the final clause is stored in the clause address storage section. Clause address 9 corresponding to “I’m arriving”
is memorized. In the case of the second sentence pattern, the first
Regarding the sentence pattern, the fixed/variable storage section from address 5 to address 8 next to the final sentence pattern address 4, the end storage section,
They are each stored in the bunsetsu address storage section. Reference numeral 4 denotes a sentence table, which sequentially stores the sentence pattern addresses and the variable number of phrases of the first sentence type and second sentence type in the sentence structure table 3, as shown in Table 4. That is, in the sentence structure table 3, the first clause of the first sentence pattern is stored at sentence pattern address 1, so 1 is stored in the sentence pattern address storage section at address 1 of the sentence table 4, and the first sentence pattern has a variable number of clauses. is 1
Therefore, 1 is stored in the variable clause number storage unit at address 1 of the sentence table 4. Also, since the first sentence pattern of the second sentence pattern is stored at address 5 of the sentence structure table 3, 5 is stored in the sentence pattern address storage section at address 2 of the sentence table 4, and the number of variable clauses of the second sentence pattern is 2. Therefore, 2 is stored in the variable clause number storage unit at address 2 of the sentence table 4.

[Table] When announcing the sentence of the first sentence type, ``The train will arrive at platform 3 soon,'' using voice memory 1, phrase table 2, sentence structure table 3, and sentence table 4. think about. Assume that the conversion switch 5 is now switched to the text table 4 side and the gate 6 is closed. Since the sentence to be announced is of the first sentence type, a signal for reading address 1, that is, a first sentence pattern reading signal, is applied to the sentence table 4 via the changeover switch 5. This causes register counter 7
The memory value "1" of the sentence pattern address storage section at address 1 of the sentence table 4 is stored in . As a result, the sentence pattern address of the sentence structure table 3 is read out. Here, the storage value of the fixed/variable storage section for sentence pattern address 1 is "0" indicating that it is a fixed phrase, so the changeover switch 5 and gate 6 maintain the current state, and the storage value of the sentence pattern address storage section is supplied to the phrase table 2 via the adder 8, and phrase address 7 is read from the phrase table 2. With this, voice memory 1
Segment numbers 0 to 3 are read out,
The phrase "soon" is read out. After this, the end storage section of sentence pattern address 1 in sentence structure table 3 is "0" indicating that there is a subsequent clause.
Therefore, the register counter 7 is updated to 2, and the sentence pattern address 2 of the sentence structure table 3 is read out. Here, since the stored value of the fixed/variable storage unit at sentence pattern address 2 is "1" representing a variable clause, the changeover switch 5 is switched to the gate 6 side, and the gate 6 is also opened. At the same time, the storage content 1 of the phrase address storage section is supplied to the adder 8. In this state, the variable phrase signal "2" is supplied to the adder 8 via the changeover switch 5 and the gate 6, and the output of the adder 8 is 3.
Then, phrase address 3 of phrase table 3 is read out, segment numbers 64 to 70 of the voice memory are read out, and the phrase "on line 3" is read out. After this, the phrases ``the train'' and ``are arriving'' are read out in the same way as the phrase ``soon.'' However, when reading out "Arrived", the end storage section of sentence pattern address 4 in sentence structure table 3 is "1" indicating that there is no subsequent phrase.
Therefore, the register counter 7 does not increment to "5" but returns to "0" and prepares for next reading. As described above, the sentence ``The train will soon arrive at platform 3.'' is read out, but if it is ``on platform 1'' instead of ``on platform 3'', ``0'' is used.
In the case of "on the second line", "1" is supplied to the adder 8 as a variable phrase signal. The same process is performed for the second sentence pattern. As described above, according to the present invention, the sentence structure table includes the sentence pattern address of each sentence pattern, information on whether the phrase at this sentence pattern address is a fixed phrase or a variable phrase, and information on whether it is the final phrase. and the clause addresses of the clauses are memorized, and the sentence pattern addresses of the beginnings of these sentence patterns are stored in the sentence table, so it is possible to store all the beginning addresses of each clause of the sentence to be announced. Compared to,
The amount of information that must be memorized is significantly reduced,
The capacity of each memory can be reduced. In the above embodiment, each segment of the voice memory 1 is made up of one sound for convenience of explanation, but originally each segment is divided into equal intervals regardless of sound and length. Further, the number of variable phrases stored in the sentence table 4 is for preventing malfunction by comparing with the generating circuit of the variable phrase signal supplied to the gate 6, although the detailed circuit configuration is omitted. Furthermore, there is silence (pause) in the sentence composition table.
It is also possible to add information regarding.

[Brief explanation of drawings]

The figure is a block diagram of a device used in the audio editing method according to the present invention. 1...Voice memory, 2...Phrase table (first
3... Sentence structure table (second memory), 4... Sentence table (third memory).

Claims (1)

[Claims] 1. A method of generating multiple types of sentences by combining fixed phrases that do not change and variable phrases that use phrases selected from a predetermined group of phrases, and which stores a large number of phrases. a voice memory; a first memory storing the storage positions of the plurality of phrases in the voice memory in such a manner that the phrase groups used for the variable phrases are located at consecutive addresses for each phrase group; For each of the above sentences, information on whether or not it is the final clause in the order of arrangement within the sentence, information on whether it is a fixed clause or a variable clause, and information on whether it is a fixed clause or a variable clause, and in the case of a fixed clause, the first clause. a second memory that stores address information that is a storage address in the memory and is a storage address in the first memory of the first clause in a group of clauses used for variable clauses, and a second memory; and a third memory storing a storage address of information about the first clause in each of the sentences, wherein information about the first clause of a predetermined sentence is stored from the third memory. A first step of reading out the address of the second memory where the phrase is located and thereby reading out the above information from the corresponding address of the second memory, and determining from the above information whether it is a fixed phrase or a variable phrase, and if it is a fixed phrase, the information is a certain address as the first
As a result, the phrase corresponding to the address of the first memory is read from the voice memory, and in the case of a variable phrase, the separately supplied numerical value is added to the address in the information by the adding means. A second step of reading out the address obtained from the first memory and thereby reading the phrase from the voice memory, and determining from the above information whether or not it is the final phrase, and if it is not the final phrase, reading the next address in the second memory. A sound editing method in a sound editing device, comprising a third step of reading out the phrase, repeating the second step, and stopping in the case of the final phrase.