 |
Description  |
|
|
BACKGROUND OF THE INVENTION
1. Field of the Invention
This invention relates to a communication method, a voice transmission
apparatus and a voice reception apparatus for use for communication
through a non-guarantee type network such as an internet.
2. Description of the Related Art
As popularization of an internet proceeds, a technique has been proposed
wherein voice is transmitted reciprocally through internet networks to
effect bidirectional communication similarly as in a public telephone
network. Such a technique of the type just described is called internet
phone service.
The internet phone service at present is influenced, because of
intervention of a number of networks each including a computer and a
router, by a delay by a load of a server at each node such that it suffers
from a delay, unnatural interruption of language, jump of voice and so
forth, and it is usually the case that complete communication cannot be
achieved. This arises from the fact that the internet is a non-guarantee
type network (also called best effort type network) which does not
guarantee time and arrival of information.
In order to solve the problem just described, a real time protocol and a
reserve protocol which assures a line have been proposed. However, they
still fail to guarantee complete communication between parties because a
network is used commonly basically by a large number of communicating
parties.
While the transmission capacity of the internet itself naturally takes part
in a cause of the problem described above, also the compression capacity
of voice data takes part in it. In compression of voice data, if the
compression ratio is raised, then the voice quality is deteriorated, but
if the compression ratio is suppressed low, then although the voice
quality is raised, since an increased band width of the transmission line
is consumed, a delay is produced and skipping, jumping or blank of voice
occurs. In either case, a desired voice quality cannot be obtained.
One of techniques for sending much voice information in a narrow band is to
convert inputted voice of a talking person into character data by a speech
recognition technique and transmit the character data to the reception
side. Since the information amount of character data is much smaller than
that of voice information, the communication delay can be reduced, and
besides, any problem involved in speech recognition does not occur. A
technique of the type described is disclosed, for example, in Japanese
Patent Laid-Open Application No. Heisei 60-136450 or Japanese Patent
Laid-Open Application No. Heisei 61-256848.
The former document discloses a system wherein input voice is recognized
first and then converted into data in the form of a packet of a packet
exchange and the data are communicated between terminals of the packet
exchange, and proposes the system as a countermeasure to improve the
processing efficiency of the exchange. In the system disclosed, since it
involves communication in the single packet exchange network, no
countermeasure is taken against a long delay or a load variation which
occurs in internet networks wherein communication is performed through a
large number of unknown nodes. Further, as recited in the document, it is
difficult to apply the system to flexible and wide range information
transmission in that reproduction is performed only with a uniform tone
and no attention is paid to natural voice.
The latter document discloses another system wherein speech recognition is
performed to obtain character codes by an originating terminal and the
character codes are sent through an exchange to a terminating terminal and
then speech synthesis is performed based on the character codes by the
terminating terminal. Since also the system involves communication in a
single communication network, similarly to the system of the former
document, no countermeasure is taken against a long delay or a load
variation which occurs in internet networks wherein communication is
performed through a large number of unknown nodes. Also, no countermeasure
is taken for real time conversion or conversion into natural voice.
Accordingly, the systems described above have the following subjects to be
solved.
The first subject resides in that conversation which allows recognition of
significance is disturbed by deterioration of voice, unnatural
interruption of language, jump of voice or the like which arises from the
transmission capacity of internet networks themselves or unstable
communication or a load variation because of intervention of an unknown
communication path such as servers provided in multiple stages.
The second subject resides in that, also in a voice transmission system
which employs speech recognition, where only transmission using character
codes is involved, since mechanical voice is reproduced, natural
conversation cannot be achieved and there is the possibility that such a
problem as misunderstanding may possibly occur.
SUMMARY OF THE INVENTION
It is an object of the present invention to provide a communication method,
a voice transmission apparatus and a voice reception apparatus by which,
in communication through a non-guarantee type network, deterioration in
voice, unnatural interruption of language, jump of voice and so forth
which arise from unstable communication or a load variation can be
eliminated to the utmost to allow conversation which can be recognized
readily and conversation by natural voice can be achieved.
In order to attain the object described above, according to an aspect of
the present invention, there is provided a communication method for
communicating voice of a talking person through a non-guarantee type
network, comprising the steps of converting, by a call originating side
terminal, voice of a talking person into electric voice data and speech
recognizing the voice data to covert the voice data into character data,
connecting the call originating side terminal to a call terminating side
terminal through the non-guarantee type network based on identification
data of the call terminating side terminal, transmitting, after the
connection between the call originating side terminal and the call
terminating side terminal is established, the character data as packets
toward the call terminating side terminal, and receiving the packet
character data by the call terminating side terminal, converting the
received character data in the form of packets into voice data by speech
synthesis and outputting the voice data as voice.
With the communication method, the following advantages can be achieved.
First, the amount of data by recognition can be reduced to eliminate a
delay in communication. Since significance is interpreted for each phrase
or group of words of talked language to convert the talked language into
character data of a form which exhibits a significantly reduced amount of
data, a delay in communication can be reduced remarkably. Consequently,
correct information can be obtained at a comparatively early timing.
Further, significance information by recognition can be transmitted with
certainty to the other party. Since talked language is recognized and
converted into character information of phrases, the information can be
reproduced not as voice which is ambiguous information which is liable to
be influenced by noise but as language having clear significance, and
consequently, a will of the talking person can be conveyed correctly.
In this instance, in order to convey correct information to the other party
of communication, an accurate sentence must be communicated even if some
delay is involved. In a conventional method, since talked language is
encoded without relying upon contents or significance of language, if some
delay or interruption occurs in or with communication, there is the
possibility that the significance cannot be conveyed at all. Therefore, as
a countermeasure to transmit voice with a higher degree of clarity and
certainty, the following features may be adopted.
In particular, in the communication method described above, the voice data
may be speech recognized divisionally for each predetermined number of
words to convert the voice data into character data. In this instance, the
voice data may be speech recognized divisionally for each phrase to covert
the voice data into character data.
Alternatively, in the communication method described above, in parallel to
the speech recognition, the speech data may be compressed in units of a
predetermined number of words to produce compression voice data, and when
the speech recognition is insufficient or results in failure, the
compression voice data may be sent out as packets.
As another alternative, the communication method may be constructed such
that a voice profile which represents a characteristic of voice inputted
to the call originating side terminal is transmitted in advance from the
call originating side terminal to the call terminating side terminal, and
upon speech synthesis by the call terminating side terminal, the character
data are converted into voice data based on the voice profile, or else
such that a speech synthesis standard pattern for standard voice is
prepared by both of the call originating side terminal and the call
terminating side terminal, and the call originating side terminal produces
finite difference data of voice inputted to the call originating side
terminal from the standard voice and sends out the finite difference data
together with the packet character data whereas the call terminating side
terminal synthesizes voice based on the speech synthesis standard pattern
and the finite difference data.
As a further alternative, the communication method may be constructed such
that the call originating side terminal sends out background sound source
data corresponding to background sound together with the packet character
data whereas the call terminating side terminal outputs the voice data and
background sound in a superposed relationship with each other, or such
that sample voice data are prepared in advance in the call originating
side terminal, and when voice of the talking person is inputted first to
the call originating side terminal, the voice data and the sample voice
data are compared with each other to discriminate whether or not the voice
data and the sample voice data originate from the same talking person, and
then only when it is discriminated that the voice data and the sample
voice data originate from the same talking person, transmission is
permitted. The communication method may otherwise be constructed such that
the call originating side terminal produces image data obtained by imaging
the talking person and sends out the image data to the call terminating
side terminal, and the call terminating side terminal displays an image
based on the received image data. Or else, the call terminating side
terminal may display the received packet character data in the form of
characters, or the call terminating side terminal may convert the received
packet character data into an electronic mail.
In summary, in the communication method described above, deterioration in
voice, unnatural interruption of language, jump of voice and so forth
which arise from unstable communication or a variation in load in a
communication method which is performed through a non-guarantee type
network can be eliminated to the utmost to allow natural conversation
which can be interpreted readily.
According to another aspect of the present invention, there is provided a
voice transmission apparatus, comprising voice inputting means for
converting voice of a talking person into electric voice data, speech
recognition means for speech recognizing the voice data to convert the
voice data into character data, identification data inputting means for
inputting identification data for identification of a particular
communication object party on a non-guarantee type network, network
connection means for connecting the voice transmission apparatus to the
particular communication object party based on the identification data
through the non-guarantee type network, and transmission means for sending
out the character data as packets to the non-guarantee type network.
The voice transmission apparatus may be constructed such that the speech
recognition means converts, if, upon recognition of the voice to convert
the voice into character data, the conversion into character data results
in success, the character data, whose conversion has resulted in success,
into packets, but if the recognition results in failure, the speech
recognition compression codes the voice data, whose recognition has
resulted in failure, and converts the compression coded voice data into
packets.
According to a further aspect of the present invention, there is provided a
voice reception apparatus, comprising network connection means for
connecting the voice reception apparatus to a non-guarantee type network
based on an identification signal which specifies the voice reception
apparatus, reception means for receiving character data in the form of
packets from the non-guarantee type network, speech synthesis conversion
means for converting the received packet character data into voice data by
speech synthesis, and voice outputting means for outputting voice based on
the voice data obtained by the conversion.
The above and other objects, features and advantages of the present
invention will become apparent from the following description and the
appended claims, taken in conjunction with the accompanying drawings in
which like parts or elements are denoted by like reference characters.
BRIEF DESCRIPTION OF THE DRAWINGS
FIGS. 1(a) is a block diagram of a communication system to which a
communication method according to the present invention is applied;
FIG. 1(b) is a similar view but showing a modified communication system;
FIG. 2 is a block diagram of another communication system to which the
communication method according to the resent invention is applied;
FIG. 3 is a block diagram of a speech recognition-synthesis conversion
section of the communication system of FIG. 2;
FIG. 4 is a flow chart illustrating operation of the communication system
shown in FIG. 2;
FIGS. 5(a) and 5(b) are flow diagrams illustrating operation of the
communication system shown in FIG. 2; and
FIGS. 6(a) and 6(b) are flow diagrams illustrating different data
transmission procedures of the communication system shown in FIG. 2.
DESCRIPTION OF THE PREFERRED EMBODIMENTS
Referring first to FIG. 1(a), there are shown a communication system
including a voice transmission apparatus and a voice reception apparatus
to which a communication method according to the present invention is
applied. In the transmission side terminal or voice transmission
apparatus, voice of a talking person is inputted through a voice to
electric conversion element such as a microphone la to a voice inputting
element 1A, by which the voice signal is converted into a corresponding
voice data electric signal. The voice data is inputted to a speech
recognition conversion section 2A, by which it is converted into a
character code data signal using speech recognition. A network
control-accessing section 4 formed from, for example, a modem or a
terminal adapter is connected to an internet 101 through a public
telephone line 51 and an internet service provider 52. After the network
control-accessing section 4 establishes connection to the network
control-accessing section 4 of a receiving side terminal of the other
party of communication based on the Internet protocol, it enters
communication in which speech recognition is used. In this instance, the
transmission side terminal knows an identification number, generally an IP
(Internet Protocol) address, of the reception side terminal in advance,
and establishes connection to the reception side terminal using the IP
address.
After the establishment of connection to the reception side terminal, a
packet transmission and communication control section 5a partitions the
character code data signal for each suitable length to form packets, adds
the IP address, a data number and so forth to the packet data and
transmits the resulting packets to the reception side terminal through the
network control-accessing section 4. In the reception side terminal or
voice reception apparatus, the received packet data are demodulated into a
character code data signal by a packet reception and communication control
section 5b and then converted into a voice data signal by speech synthesis
by a speech synthesis conversion section 2B. Then, a voice outputting
element 1B outputs the voice data signal as voice by means of an electric
to voice conversion element such as a speaker 1b.
FIG. 1(b) shows in block diagram a modification to the communication system
to which the communication method according to the present invention is
applied.
Referring to FIG. 1(b), in the modified system shown, in a condition
wherein the transmission side terminal and the reception side terminal are
connected to an internet phone service provider 102 in the internet 101,
an identification number sent out from the transmission side terminal is
received by the internet phone service provider 102, and then the internet
phone service provider 102 executes a connection process to the reception
side terminal, whereafter communication based on speech recognition is
started between the terminals.
This connection is applied to a case wherein the other party terminal does
not have a fixed identification number (IP address) in such a case that
the other party terminal is connected to the internet service provider 52
by dial up connection. In this instance, if identification ID information
necessary for connection such as telephone numbers or mail addresses of
the transmission and reception side terminals are registered in advance in
the Internet phone service provider 102 and the transmission side terminal
inputs the identification ID information of the reception side terminal,
then the internet phone service provider 102 originates, based on the
identification ID information, a call to the reception side terminal
through an accessing device such as a modem or a terminal adapter to
establish connection to the reception side terminal. Or, also where the
two terminals are connected to the internet phone service provider 102
simultaneously, establishment of connection is possible.
The systems shown in FIGS. 1(a) and 1(b) can be constructed in a type
wherein a communication apparatus is added to or combined with a
processing apparatus such as a personal computer (PC) or another type
wherein a processing section and a communication section are formed as a
unitary member or in a further type wherein characteristics of the two
types are combined.
FIG. 2 shows in block diagram a communication system including a voice
transmission apparatus and a voice reception apparatus to which the
communication method according to the present invention is applied.
Referring to FIG. 2, each of a transmission side terminal 20a and a
reception side terminal 20b includes a voice inputting and outputting
element (voice inputting means and voice outputting means) 1 having a
microphone la and a speaker 1b built therein, a speech recognition and
synthesis conversion section (speech recognition means and speech
synthesis conversion means) 2 for recognizing voice data from the voice
inputting and outputting element 1 by speech recognition and converting
the voice data into character code data, a communication control section
(transmission means and reception means) 3 for controlling an entire
communication control procedure, a network control apparatus (network
connection means) 4, a modem 5, an inputting operation section 6 for
inputting various data therethrough, and a voice or data storage and
display section 7.
Referring now to FIG. 3, the speech recognition and synthesis conversion
section 2 includes a speech recognition circuit 2a, a compression coding
circuit 2b, and a selection circuit 2c. The speech recognition circuit 2a
compares speech recognized voice data with terms of a registration
dictionary stored in advance therein to recognize the voice data in units
of a phrase (group of words which may include a single word) like "hello",
and outputs a result of the recognition as character codes to the
selection circuit 2c. Also a signal representative of whether or not the
recognition results in success or failure (the recognition result is not
any one of voices registered in the registration dictionary) is outputted
to the selection circuit 2c. The compression coding circuit 2b compression
codes the voice signal and outputs resulting codes to the selection
circuit 2c. The selection circuit 2c normally selects the output of the
speech recognition circuit 2a, but if a signal representing that the
recognition results in failure is received, then the selection circuit 2c
selects the output of the compression coding circuit 2b.
Based on the identification number of the reception side terminal 20b
inputted from the inputting operation section 6 of the transmission side
terminal 20a, the transmission side terminal 20a is connected to an
internet 101 through a telephone line 51 and an internet service provider
(ISP) 52. Then, by a similar procedure, the transmission side terminal 20a
establishes connection to the reception side terminal 20b connected to the
Internet 101 based on the internet protocol. Then, upon transmission of
voice, the voice is recognized and converted into character codes by the
speech recognition and synthesis conversion section 2, and the character
codes are converted into packets and the IP address is added to the
packets, whereafter the packets are sent to the reception side terminal
20b by a TCP/IP. The reception side terminal 20b converts the character
codes into voice by speech synthesis to convey the voice to the other
party. On the other hand, upon reception of voice, packet data sent from
the other party terminal 20b are successively synthesized into voice by
the transmission side terminal 20a. Consequently, the talking persons can
talk in continuous sentences with each other.
Operation of the communication system having the construction described
above is described in detail with reference to FIG. 2. The voice inputting
and outputting element 1 of the transmission side terminal 20a first
converts voice inputted from the microphone la from an analog signal into
a digital signal and sends the digital signal to the speech recognition
and synthesis conversion section 2. The speech recognition and synthesis
conversion section 2 converts the input voice into character codes using
the voice pattern and the registration dictionary. After the inputting is
completed, the communication control section 3 immediately converts the
character codes into packets, adds the IP address to the packets and
transfers the resulting packets to the other party terminal through the
modem 5 and network control-accessing section 4 of the transmission side
terminal 20a and the communication line.
Here, the voice is divided for individual words and thus identified
contents are sent to the other party terminal after a minimum delay.
However, since the identified contents are conveyed to the other party
through the internet service provider 52 and the internet 101 connected to
each other, they arrive at the other party after a delay of one breath
because of an influence of such a delay time of the network.
It is to be noted that the speech recognition and synthesis conversion
section 2 shown in FIG. 2 has not only the voice text conversion function
but also functions of format conversion, compression coding and so forth.
FIG. 4 illustrates, in flow chart, detailed operation of the voice
transmission apparatus and the voice reception apparatus of the
communication system shown in FIG. 2.
Referring to FIG. 4, internet call connection from the call originating
side (transmission side) terminal 20a by dial up or the like to the call
terminating side (reception side) terminal 20b is performed.
In the internet call connection, the network control section 4 of the call
originating side terminal 20a logs on to the internet service provider 52
through the public telephone line 51 and connects to the internet 101.
Then, the internet phone service provider 102 is connected from the
communication control section 3 through the network control section 4 and
the modem 5, and an identification number of the other party to be
connected, that is, the call terminating side terminal 20b, is sent out.
The internet phone server 102 has a table for conversion between
identification numbers and the other party IP addresses and sends a
request for connection to the call terminating side terminal 20b making
use of the IP address of the call terminating side terminal 20b detected
from the table. It is assumed that either the call terminating side
terminal 20b logs on to the internet 101 by dial up or the like in advance
under the control of the internet phone service provider 102 and connects
to the internet phone service provider 102 or the internet phone service
provider 102 performs telephone connection through an access point nearest
to the call terminating side terminal 20b and the call terminating side
terminal 20b completes its telephone call connection and internet network
connection by automatic response or the like.
The internet call connection is performed by a method wherein it is
performed by intervention of such a mechanism which effects re-reading
between an IP address and an individual name such as the internet phone
service provider 102 as described above or another method wherein the
other party has a fixed IP address and the call origination side knows the
IP address in advance and sends out the IP address.
The procedure described above is executed in steps S101 and S201 of FIG. 4.
After completion of the internet call connection, conversation is performed
in a procedure described below.
Conversation with the other party is performed using the speech recognition
and compression and decompression functions of the speech recognition and
synthesis conversion section 2 of the terminals 20a and 20b. In this
instance, speech recognition is performed principally, and the compression
and decompression functions are used preparatorily.
Prior to conversation, the communication control section 3 of the call
originating side terminal 20a confirms by a protocol with the call
terminating side terminal 20b whether or not data and a voice profile for
allowing, upon speech synthesis and reproduction of the other party side,
reproduction with a tone near to the pronunciation of the call originating
party should be sent out. If a result of the confirmation reveals that "to
be sent" has been selected, the a standard voice profile of the call
originating side terminal 20a, or particularly when it is desired to
change the tone, a voice profile designated particularly, is sent out to
the other party terminal 20b. Similarly, confirmation of transmission of a
voice profile is performed also from the call terminating side terminal
20b, and when "to be sent" is selected, a voice profile is sent from the
call terminating side terminal 20b and is received by the call originating
side terminal 20a. The voice profiles thus sent are subject to
incorporation operation by both of the call terminating and originating
sides. A voice profile includes rules (for example, voice of a man or a
woman) to be used for conversion from characters into voice, which
characterize a tone of a talking person, indicated by digital patterns of
acoustic waves and is obtained by learning by the transmission side
terminal 20a in advance. Therefore, if the character code data sent with
such patterns incorporated therein are synthesized into voice by the
speech recognition and synthesis conversion section 2 of the call
terminating side terminal 20b, then reproduction voice close to the tone
of the talking person is obtained.
The procedure described above is executed in steps S102 to S105 and S202 to
S205 of FIG. 4.
Subsequently, a procedure of the transmission side terminal 20a in a stage
after actual conversation is entered is described in connection with steps
S106 to S110.
In the conversation, the following operations are repeated to convey
language to the other party. In particular, uttered voice is subject to
cutting out of voice data in units of a phrase, speech recognition and
confirmation of success/failure of recognition by the speech recognition
circuit 2a of the speech recognition and synthesis conversion section 2.
If the confirmation proves success, then the character code data after the
recognition are sent as packet data to the call terminating side terminal
20b through the selection circuit 2c. Since speech recognition does not
exhibit a recognition rate of 100%, a case wherein recognition results in
failure due to ambiguous pronunciation or by some other reason occurs at a
certain rate. For preparations for a case of failure, also compression
data cut out at the same position as that in the speech recognition are
produced parallelly by the compression coding circuit 2b, and if the
speech recognition results in failure, then the data part with which the
recognition has resulted in failure is filled not with the recognized data
but with the compression data.
"Cutting out" of voice is normally performed in units of an ordinary
phrase, In particular, voice is successively recognized in units like
"Hello", "I am" or the like.
Then, the speech recognition circuit 2a compares the uttered voice with the
registration dictionary and adopts a pattern having a comparatively high
similarity. Since any data with which no similar word has been found as a
result of the comparison is either a new word or an unrecognizable word,
the voice compression data produced by the compression coding circuit 2b
is used to fill by the selection circuit 2c.
Subsequently, a procedure of the call terminating side terminal 20b in a
stage after actual conversation is entered is described in connection with
steps S210 to S214.
The call terminating side terminal 20b checks packet data sent thereto from
the call originating side terminal 20a. If the packet data are data of the
text type, that Is, character code data, then the speech recognition and
synthesis conversion section 2 converts them into voice data using voice
patterns based on the voice profiles incorporated already therein and
sends out the voice data to the voice inputting and outputting element 1
which includes the speaker 1b. On the other hand, if the packet data are
compression data, then the speech recognition and synthesis conversion
section 2 converts the compression data into voice data corresponding to
the input voice using the decompression function thereof and sends out the
voice data to the voice inputting and outputting element 1.
Subsequently, in what manner talked language is reproduced by the reception
side apparatus is described with reference to FIGS. 5(a) and 5(b) which
are time charts illustrating the relationships in time of data processing.
Where the talked language is, for example, "Hello! This is Tanaka. Please
call Tashiro.", it is divided into four phrases of "Hello", "This is
Tanaka", "Please call" and "Mr. Tashiro". Thus, the phrases are
successively recognize | | |
