JP6648786B2 - Voice control device, voice control method and program - Google Patents

Description

  The present invention relates to a voice control device, a voice control method, and a program.

In recent years, the following has been proposed as a speech synthesis technique. That is, by responding with a voice synthesized according to the tone and voice quality of the user, a technique of pronouncing more humanly (for example, see Patent Document 1), analyzing the voice of the user, and analyzing the psychology of the user A technology for diagnosing a condition or a health condition (for example, see Patent Document 2) has been proposed.
Further, while recognizing a voice input by a user, a voice dialogue system that realizes voice dialogue with the user by outputting the content specified in the scenario by voice synthesis has been proposed (for example, Patent Document 3). reference).

JP 2003-271194 A Japanese Patent No. 4495907 Japanese Patent No. 4832097

By the way, it is assumed that a dialogue system that combines the above-described voice synthesis technology and a voice dialogue system to search for data and output by voice synthesis in response to a user's voice question. In this case, a problem has been pointed out that the voice output by the voice synthesis gives the user an unnatural feeling, specifically, a feeling as if the machine is talking.
The present invention has been made in view of such circumstances, and one of its objects is to provide a user with a natural feeling, specifically, to give a user a psychologically good impression. The present invention provides a voice control device, a voice control method, and a program that can give a user an impression as if interacting with a person.

  In examining a man-machine system that outputs (answers) a response to a question by a user through speech synthesis, the present inventor first determines the pitch (frequency ).

  Here, as a dialogue between persons, a case will be examined in which a question (question) by one person (a) is answered by the other person (b). In this case, when “a” asks a question, not only “a” but also “b” who tries to answer the question often leaves a strong pitch in a certain section of the question. b, when responding with the meaning of consent, agreement, affirmation, etc., the part that characterizes the answer, such as the pitch of the ending or initial, has a predetermined relationship with the pitch of the question remaining in the impression. Reply to When listening to the answer a, the pitch remaining in the impression of the self question and the pitch characterizing the answer to the question have the above relationship. The inventor of this case thought that it would have a good impression.

  For example, when a asks a question, "Are you sure?", A and b remember the pitch of the end of the question, "sho", which has strong meaning such as reminding or confirmation. State. In this state, when b tries to answer “a, yes” affirmatively to the question, the part characterizing the answer to the pitch of “sho” remaining in the impression, for example, the ending "Yes," so that the pitch of "Yes" has the above relationship.

FIG. 2 shows a formant in such an actual dialog. In this figure, the horizontal axis represents time, the vertical axis represents frequency, and the spectrum shows a state in which the intensity increases as the color becomes whiter.
As shown in the figure, a spectrum obtained by frequency-analyzing a human voice appears as a plurality of temporally moving peaks, that is, formants. In detail, the formant corresponding to "Yes?" And the formant corresponding to "Oh, yes" respectively appear as three peak bands (white band-like portions moving along the time axis).
Focusing on the lowest frequency of these three peak bands, the frequency of the code A (the center part) corresponding to the sound “sho” of “yes?” Is about 400 Hz. On the other hand, the frequency of the code B corresponding to the sound "i" of "a, yes" is approximately 260 Hz. Therefore, it can be seen that the frequency of the code A is almost 3/2 of the frequency of the code B.

  The relationship that the frequency ratio is 3/2 may be expressed in terms of pitch as "do" having the same octave as "so" or "la" as one octave below "mi". , As described later. Further, this frequency ratio (predetermined relationship between pitches) is a preferred example, but various examples are given as described later.

  FIG. 3 is a diagram showing a relationship between a pitch name (floor name) and a frequency of a human voice. In this example, the frequency ratio based on the fourth octave “do” is also shown, and “so” is 3/2 as described above, based on “do”. The frequency ratio based on the third octave “La” is also illustrated in parallel.

As described above, in a dialogue between people, the pitch of a question is an element that characterizes the impression of the question. On the other hand, it is empirically recognized that the way of answering differs depending on the linguistic content of questions and answers.
For example, if you answer "Yes" to a question, you respond relatively quickly to the question, but if you answer "No", for example, consider the impression of the other party. This is the point of responding with a pause.
Therefore, when examining a dialogue system for outputting (replying) the answer to a question by a user using speech synthesis, the speech is synthesized not only for the pitch of the question but also for the linguistic meaning of the question and the answer. Can be an important factor in
Therefore, in order to achieve the above-mentioned object in the voice synthesis, the following configuration is adopted.

That is, in order to achieve the above object, a voice synthesis device according to one embodiment of the present invention includes a voice input unit that inputs a question based on a voice signal, and a sound that analyzes a pitch of a specific first section in the question. A high-analysis unit, an acquisition unit that acquires an answer to a question, a language analysis unit that analyzes linguistic information of the question and the answer, a speech synthesis unit that speech-synthesizes the obtained answer, and speech synthesis by a speech synthesis unit. A voice control unit that changes a pitch of a specific second section in the answer to a pitch having a predetermined relationship with a pitch of the first section, and controls the pitch according to language information of the question and the answer; And characterized in that:
In this aspect, the pitch of the specific second section in the answer is changed so as to have a predetermined relationship with the pitch of the specific first section of the question, so that the voice is synthesized. The answer can give the user a good impression of comfort and security. In addition, since speech synthesis is controlled in accordance with the language information of the question and the answer, it is possible to give the user an impression as if he or she is interacting with a person.

In this aspect, the first section is preferably, for example, the ending of the question, and the second section is preferably the beginning or ending of the answer. As described above, the section characterizing the impression of the question is the ending of the question, and the section characterizing the impression of the answer is often the beginning or ending of the answer.
Further, the predetermined relationship is preferably a relationship of consonance intervals excluding a perfect degree. Here, the consonance refers to a relationship in which, when a plurality of musical tones are generated simultaneously, they are fused and well harmonized, and these pitch relationships are referred to as a consonance interval. The degree of consonance is higher as the frequency ratio (frequency ratio) between two sounds is simpler. The simplest of the frequency ratios, 1/1 (perfect 1 degree) and 2/1 (perfect 8 degrees), are particularly called absolute consonance pitches, which are 3/2 (perfect 5 degrees) and 4/3 (perfect). (4th degree) plus a complete consonance. 5/4 (major third), 6/5 (minor third), 5/3 (major sixth) and 8/5 (minor sixth) are called incomplete consonance intervals, and all other frequency ratios (Long and short second and seventh degrees, various increasing and decreasing pitches, etc.) are called dissonant pitches.

Note that if the pitch of the beginning or end of the answer is the same as the pitch of the end of the question, it is considered that the dialogue is accompanied by an unnatural feeling. Has been removed.
In the above aspect, the most preferable example of the predetermined relation is considered to be a relation in which the pitch of the second section is a consonant pitch 5 degrees lower than the pitch of the first section as described above. However, the predetermined relationship is not limited to a consonant pitch except for a perfect degree, and may be a dissonant pitch or a pitch relationship in the range of one octave above and below, excluding the same.
In addition, the answer is not limited to a concrete answer to the question, but also includes a hammer (interjection) such as "I see,""Isee."

Further, in a dialogue between people, for example, when the answer to “Yes” is given to a question that requires a choice between “Yes” and “No”, the pitch is relatively high, while “No” Reducing the pitch when answering is a common behavior in experience. Therefore, in the above aspect, the voice control unit may have a configuration in which the pitch of the positive answer to the question is higher than the pitch of the negative answer. According to this configuration, it is possible to create an atmosphere for the user as if the user is interacting with each other.
In addition, affirmative answers include consent such as "yes" and "yes" in addition to "yes", and negative answers include "unhappy" and "agree" in addition to "no". I can't agree. "

  By the way, when answering a question in a dialogue between people, there are factors to be considered other than the pitch. One of these elements is between the time the question is asked and the time the answer is output. For example, when answering "No" to a question that requires a two-part answer as described above, an act that is often seen in experience is that taking a delay to take a pause to take care It is. Therefore, the voice control unit may control the voice synthesis unit so that the time from when the question is issued until the output of the positive answer is shorter than that between the negative answers.

In a dialogue between people, instead of two choices, for example, 5W1H such as Who (who), What (what), When (when), Where (where), Why (why), How (how). In some cases, specific questions are answered slowly over time. On the other hand, even in the speech synthesizer, it sometimes takes time to obtain an answer of the specific content.
In any case, if there is not enough time between the question and the answer to the specific content, the questioning side (user) is given a kind of anxiety and the subsequent conversation does not bounce.
On the other hand, if the time until the answer is too short, do you not feel as if you are consciously covered or do you not listen to people's story properly? , Feeling uncomfortable.
Therefore, in the above aspect, when the language information of the question has predetermined content, the voice synthesis unit voice-synthesizes an answer with predetermined content, and the voice control unit transmits the answer after the question is issued. The configuration may be such that the speech synthesis unit is controlled so that the time until output is within a predetermined time range.
According to this configuration, the answer of the predetermined content is voice-synthesized within the predetermined time before the answer of the specific content to the question, so that the user does not feel uneasy. it can.
In addition, as the answer of the predetermined content, for example, an interjection such as “um” or “yes” is given. Further, it is considered that a time range from 0.5 seconds to 2.0 seconds is appropriate as the range of the predetermined time.

The aspects of the present invention can be conceptualized not only as a speech synthesizer but also as a program that causes a computer to function as the speech synthesizer.
In the present invention, the pitch (frequency) of the question is analyzed and the pitch of the answer is controlled. However, as is clear from the above-described example of the formant, the human voice has a certain frequency range. Therefore, it is inevitable that analysis and control have a certain frequency range. In addition, errors occur in analysis and control as a matter of course. For this reason, in the present case, regarding the analysis and control of the pitch, not only that the numerical value of the pitch (frequency) is the same, but also that a certain range is allowed.

It is a block diagram showing the composition of the speech synthesis device concerning an embodiment. It is a figure showing the example of the voice formant in a dialog. FIG. 4 is a diagram illustrating a relationship between a pitch name, a frequency, and the like. 5 is a flowchart illustrating an operation of the speech synthesis device. It is a figure which shows the specific specific example of a ending. FIG. 4 is a diagram illustrating an example of pitch shift with respect to a voice sequence. FIG. 4 is a diagram illustrating an example of pitch shift with respect to a voice sequence. FIG. 4 is a diagram illustrating an example of pitch shift with respect to a voice sequence. It is a figure explaining output timing of a reply. It is a figure explaining output timing of a reply. It is a figure showing the important section of processing in an example of application (the 1). It is a figure showing an important section of processing in an example of application (the 2). It is a figure showing an important section of processing in an example of application (the 3).

Hereinafter, embodiments of the present invention will be described with reference to the drawings.
FIG. 1 is a diagram illustrating a configuration of a speech synthesis device 10 according to the embodiment. In FIG. 1, a voice synthesizer 10 is a terminal device such as a mobile phone having a voice input unit 102 and a speaker 142, and a plurality of functional blocks are implemented by executing a pre-installed application program by a CPU. It is constructed as follows. In detail, in the speech synthesizer 10, the utterance section detection unit 104, the pitch analysis unit 106, the language analysis unit 108, the speech control unit 109, the answer creation unit (acquisition unit) 110, the speech synthesis unit 112, the language database 122, The answer database 124, the information acquisition unit 126, and the audio library 128 are constructed.
Although not shown, a display unit and an operation input unit are also constructed in addition to the above, so that the user can check the status of the device and input various operations to the device. It has become. The speech synthesizer 10 may be a notebook or tablet personal computer.

Although not shown in detail, the audio input unit 102 includes a microphone that converts audio into an electric signal, an LPF (low-pass filter) that cuts high-frequency components of the converted audio signal, and a high-frequency component that has been cut. An A / D converter for converting an audio signal into a digital signal.
The utterance section detection unit 104 processes an audio signal converted into a digital signal to detect an utterance (voiced) section. Specifically, the utterance section detection unit 104 detects, for example, a section in which the state in which the amplitude is equal to or smaller than the threshold continues for a predetermined period or more in the audio signal as a silent section, and detects other sections as utterance sections.

  The pitch analysis unit 106 performs a frequency analysis on the voice signal detected as the utterance section, obtains a pitch of a specific section (first section) among the first formants obtained by the analysis, and determines the pitch. Outputs pitch data indicating the pitch. The first section is, for example, the ending of the question. In addition, the first formant refers to a component having the lowest frequency among a plurality of formants obtained when, for example, a voice is subjected to frequency analysis. In the example of FIG. Say. For frequency analysis, FFT (Fast Fourier Transform) and other known methods can be used. An example of a specific method for specifying the ending in the question will be described later.

The language analysis unit 108 determines which phoneme is close to the speech signal detected as the utterance section by referring to a phoneme model created in advance in the language database 122, and determines a word (question) defined by the speech signal. ) Is analyzed (specified). Note that, for example, a hidden Markov model can be used as such a phoneme model.
Further, the language analysis unit 108 analyzes not only the question by the voice signal but also the meaning of the answer by the answer creation unit 110 described later.
The linguistic analysis unit 108 supplies the analysis result of the meaning of the question to the answer creation unit 110, and when receiving the answer from the answer creation unit 110, supplies the analysis result of the answer to the voice control unit 109.

The answer creating unit 110 creates an answer corresponding to the meaning of the word analyzed by the language analyzing unit 108 with reference to the answer database 124 and the information acquiring unit 126.
In this embodiment, the answer created by the answer creating unit 110 includes:
(1) Answers indicating affirmation, etc.
(2) Answers indicating negative or other means,
(3) Specific answers to questions,
(4) Sympathetic answers (including a hammer) issued prior to the specific answer,
Is assumed. Examples of the answer (1) include “Yes” and “Yes”, and examples of the answer (2) include “No” and “No”. As (3), for example, there is an example in which the question "What is tomorrow's weather? (What is the weather tomorrow?)" As (4), there are "Um" and "Yes" before the answer "Harere".
Regarding the answers (1) and (2), for example, in response to a question “about 3:00?”, If time information is acquired from a built-in real-time clock (not shown), the answer creation unit 110 For example, it is possible to determine which of “Yes” and “No” should be answered.
On the other hand, for example, the question “Tomorrow is fine (tomorrow is fine)?” Cannot be answered by the voice synthesizer 10 alone unless the external server is accessed to obtain weather information. As described above, when the answer cannot be made only by the speech synthesizer 10, the information acquiring unit 126 accesses the external server via the Internet, acquires information necessary for creating an answer, and supplies the information to the answer creating unit 110. The answer creating unit 110 can determine whether or not the question is correct and give an answer.
Regarding the answer (3), for example, in response to the question “What is the time? (What time is it?),” The answer creating unit 110 acquires the time information and also stores information other than the time information in the answer database. By obtaining the answer from 124, it is possible to create a reply “It is now XX hours XX minutes”. On the other hand, in response to the question "What is the weather of tomorrow? (What is the weather of tomorrow?)", The information acquisition unit 126 accesses an external server to acquire information necessary for an answer, and an answer creation unit. 110 is configured to create an answer to the question, for example, “Hello” from the answer database 124 and the external server.

  The answer creating unit 110 supplies the created and obtained answer to the language analyzing unit 108, and creates and outputs a speech sequence from the answer. This voice sequence is a phoneme sequence and defines a pitch and a sounding timing corresponding to each phoneme. For (1), (2), and (4), for example, a voice sequence corresponding to the answer may be stored in advance in the answer database 124, while a voice sequence corresponding to the answer may be read from the answer database 124. More specifically, the answer creating unit 110 may read out a voice sequence such as “Yes” or “No” according to the determination result in the answers (1) and (2), and (4) In (3), before outputting a voice sequence corresponding to the answer of (3), any one of predetermined voice sequences such as “Um” and “Oh well” may be read. The voice sequence is supplied to the voice control unit 109 and the voice synthesis unit 112, respectively.

  The voice control unit 109 determines control content for the voice sequence according to the pitch data supplied from the pitch analysis unit 106 and the meaning of the answer supplied from the language analysis unit 108. The contents of control by the audio control unit 109 will be described later.

The voice synthesizer 112 synthesizes voice according to a voice sequence of an answer in which a pitch and a sound generation timing are specified. Therefore, the voice synthesis unit 112 can output the basic voice of the answer by simply synthesizing the voice according to the voice sequence.
However, in the present embodiment, the voice synthesis unit 112 synthesizes the voice by changing the basic voice defined by the voice sequence according to the control content of the voice control unit 109.
Here, the voice control unit 109 controls the voice synthesis by the voice synthesis unit 112 as follows. That is, the voice control unit 109 changes the pitch of a specific section (second section) of the voice sequence so as to have a predetermined relationship with the pitch data, and generates the sound sequence of the voice sequence. The timing is determined according to the semantic content of the analysis result of the answer by the language analysis unit 108.
In the present embodiment, the second section is set as the ending of the answer, but is not limited to the ending as described later. Further, in the present embodiment, the pitch having a predetermined relationship with the pitch data is defined as a pitch having a relationship lower by 5 degrees, but a pitch having a relationship other than 5 degrees lower as described later. It may be high.

In synthesizing speech, the speech synthesizing unit 112 uses speech unit data registered in the speech library 128. The speech library 128 is a database in which speech segment data defining waveforms of various speech segments as speech materials, such as a single phoneme or a transition from a phoneme to a phoneme, is prepared in advance. Specifically, the speech synthesis unit 112 combines the speech unit data of each sound (phoneme) of the speech sequence, and corrects the connected portion to be continuous, while changing the pitch of the answer as described above. Modify and generate an audio signal.
The audio signal synthesized by the audio synthesizing unit 112 is converted into an analog signal by a D / A conversion unit (not shown), and is then acoustically converted by the speaker 142 and output.

Next, the operation of the speech synthesizer 10 will be described.
FIG. 4 is a flowchart illustrating a processing operation in the speech synthesis device 10.
First, when the user performs a predetermined operation, for example, when an icon or the like corresponding to the process for the conversation is selected on a main menu screen (not shown), the CPU starts an application program corresponding to the process. By executing this application program, the CPU constructs the functional blocks shown in FIG.

  First, a user inputs a question by voice to the voice input unit 102 (step Sa11). The utterance section detection unit 104 detects the utterance section by comparing the amplitude of the voice with the threshold, and supplies the voice signal of the utterance section to each of the pitch analysis unit 106 and the language analysis unit 108 (step Sa12). .

The pitch analysis unit 106 analyzes the voice signal of the question in the detected utterance section, specifies the pitch of the first section (ending) in the question, and converts the pitch data indicating the pitch into the voice control unit. 109 (step Sa13). Here, an example of a specific method of specifying the ending of the question in the pitch analysis unit 106 will be described.
Assuming a dialogue in which a person who has asked a question wants an answer to the question, it is considered that the volume corresponding to the ending of the question temporarily increases in comparison with other portions. Therefore, the pitch of the first section (ending) by the pitch analyzing unit 106 can be obtained, for example, as follows.
First, the pitch analysis unit 106 separates a sound signal of a question detected as an utterance section into a sound volume and a pitch (pitch) and forms a waveform. FIG. 5A is an example of a volume waveform in which the vertical axis represents the volume of the audio signal and the horizontal axis represents the elapsed time, and FIG. 5B is the first waveform obtained by performing frequency analysis on the same audio signal. It is a pitch waveform in which the pitch of the formant is represented on the vertical axis and the elapsed time is represented on the horizontal axis. The time axis of the volume waveform of (a) and the pitch waveform of (b) are common.
Second, the pitch analysis unit 106 specifies the timing of the temporally last maximum P1 in the volume waveform (a).
Third, the pitch analysis unit 106 recognizes that a predetermined time range (for example, 100 μsec to 300 μsec) including the timing of the specified maximum P1 before and after is the ending.
Fourth, the pitch analysis unit 106 outputs, as pitch data, the average pitch of the section Q1 corresponding to the recognized ending in the pitch waveform of (b).
By specifying the last maximum P1 of the volume waveform in the utterance section as the timing corresponding to the ending of the question in this way, it is considered that erroneous detection of the ending of the question as a conversation can be reduced.

  In this case, in the volume waveform of (a), a predetermined time range including the timing of the last local maximum P1 before and after is determined to be the ending, but the predetermined timing with the timing of the local maximum P1 as the start or end is determined. May be recognized as the ending. In addition, instead of the average pitch of the section Q1 corresponding to the recognized ending, the pitch at the start and end of the section Q1 and the timing of the maximum P1 may be output as the pitch data.

On the other hand, the language analysis unit 108 analyzes the meaning of the word (question) defined by the voice signal, and supplies data indicating the meaning to the answer creating unit 110 (step Sa14).
The answer creating unit 110 creates an answer corresponding to the analyzed word (question) by using the answer database 124 and, if necessary, the information acquiring unit 126 (step Sa15), and converts the answer into a linguistic analysis. To the unit 108.
Here, the created answers (read from the answer database 124) are the answers (1), (2), and (4) to the question. Although the specific answer of (3) is omitted in this flowchart, it is created after the answer of (4).
In addition, the answer creating unit 110 outputs a sound sequence of the created and obtained answer (Step Sa16).

FIG. 6A is an example of the voice sequence of the answer “yes” indicating affirmative in (1). In the example of this figure, a note is assigned to each sound of the answer "yes" to indicate the pitch and the sounding timing of each word (phoneme). In this example, one note is assigned to one note (phoneme) for simplification of description, but a plurality of notes may be assigned to one note such as a slur or a tie.
FIG. 7A shows an example of the voice sequence of the answer “No” indicating the negative of (2), and FIG. 8A shows the voice sequence of the answer “Eto” of (4). This is an example.

The language analyzer 108 analyzes the meaning of the answer by the answer generator 110 (step Sa17). In the present embodiment, it is specified which of the above (1), (2) and (4) is the answer, and information indicating the specified result is supplied to the voice control unit 109.
The voice control unit 109 responds to the pitch data supplied from the pitch analysis unit 106 such that the pitch at the end of the voice sequence output from the answer generation unit 110 has the following relationship. (Step Sa18).

More specifically, the voice control unit 109 determines that the answer at the end of the answer is, for example, 5 degrees lower than the pitch indicated by the pitch data for the answer indicating affirmative in (1). , To shift the pitch of the entire voice sequence of the answer.
On the other hand, the voice control unit 109 determines that the answer indicating negative (2) is such that the pitch at the end of the answer is 6 degrees lower than the pitch indicated by the pitch data. Is determined to shift the pitch of the entire voice sequence. That is, the pitch of the answer (1) to the question is determined to be higher than the pitch of the negative answer (2).
Note that the voice control unit 109 sets the ending pitch of the answer to be 5 degrees lower than the pitch indicated by the pitch data, for example, as in the answer of (1), for the answer of (4). Then, it is determined that the pitch of the entire voice sequence of the answer is shifted.

Specifically, as shown in FIG. 6 (b), the pitch of the ending “re” indicated by the symbol A in the question “Tomorrow is fine?” When it is indicated that the sound sequence is “mi”, for the sound sequence “yes” of the answer indicating affirmativeness in (1), the sound control unit 109 sets the pitch “ye” at the end of the character B The pitch shift defined by the voice sequence is determined so that “” is “la” that is five degrees below “mi”.
On the other hand, as shown in FIG. 7B, for the answer “No” indicating the negative of (2) to the same question, the voice control unit 109 sets the ending “E” indicated by the code C to “No”. The pitch shift defined by the voice sequence is determined so that the pitch becomes “G” which is 6 degrees lower than “MI”.
In addition, as shown in FIG. 8 (b), in the question "What is Asuno Tenki?", The pitch of the ending "ha (wa)" indicated by the symbol A is "mi" according to the pitch data. Is indicated, the voice control unit 109 sets the pitch of the ending “to” indicated by the symbol D to 5 degrees with respect to “mi” for the voice sequence of “E-to” in the answer of (4). The pitch shift defined by the voice sequence is determined so as to be “la” in the following relationship.

On the other hand, the voice control unit 109 determines the output timing of the voice sequence, that is, the interval, according to the result of specifying the meaning of the answer (step Sa19). Specifically, the voice control unit 109 determines that the answer to the affirmative in (1) is, for example, in the range of 0.5 seconds to 2.0 seconds as the interval from the end of the question to the output of the answer. The answer to the negative in (2) is set later than the answer in (1), for example, about 3.0 seconds. As a result, a positive answer in (1) is output earlier in time than a negative answer in (2).
Note that the voice control unit 109 sets the answer in (4) as the interval, for example, in a range from 0.5 seconds to 2.0 seconds as in (1).

The voice control unit 109 controls the voice synthesizing unit 112 so that the voice sequence from the answer creating unit 110 is shifted to the determined pitch and output at the determined timing (step Sa20). In accordance with this control, the voice synthesizer 112 changes the pitch of the voice sequence, voice-synthesizes the answer, and outputs it.
Regarding the output timing of the answer, the voice control unit 109 determines that the time measured by the real-time clock (not shown) from the end of the question, for example, the timing of the maximum P1 described above, What is necessary is just to control the speech synthesis unit 112 so as to start outputting the speech sequence.
After that, although not shown, if the answer of (4) is output, the specific answer of (3) is output after the output of the answer of (4). Thereby, the execution of the application program ends, and the processing procedure returns to the menu screen.

FIG. 9 is a diagram for explaining the output of the answer of (1) or (2) to the question asked by the user. In response to the question "Tomorrow?" This is an example of a case where the speech synthesis device 10 as a terminal device outputs a response of “” or “No”.
In the example of this figure, speech synthesis is performed so that the pitches of the endings of “Yes” and “No” have a relation of a consonant pitch 5 degrees lower and 6 degrees lower than the pitch of the ending of the question. Thus, it is possible to give the impression that the user is interacting with the person without giving the user an unnatural feeling.
In the example of the figure, the pitch corresponding to the ending of “Yes” in (1) and the pitch corresponding to the ending of “No” in (2) The relationship is as described above with reference to FIGS. 6 and 7, and the pitch of the answer (finish) of “Yes” of affirmative is referred to as “No” of negative. The pitch of the answer (the end of the answer) becomes lower.
For this reason, according to the present embodiment, when responding in a negative sense, the answer is output in a low voice with consideration given not to hurt the opponent's mood. Such an impression can be given to the user.
When outputting the negative answer of (2), the voice control unit 109 may set the volume of the answer to be lower (lower) than the volume of the positive answer of (1). .

  Also, in this figure, the timing at which the question is issued, for example, the timing of the ending of the question "re" is defined as (a), the timing at which the beginning of the answer of "yes" is generated is defined as (b), and the "no" (C) is the timing at which the beginning of the answer is issued. In this example, the period T1 from when the question is issued to when the answer of "Yes" is output is earlier than the period T2 until the answer of "No" is output. That is, T1 <T2, in other words, in the speech synthesizer 10 according to the embodiment, when the answer is “No”, the speech is output one breath later than when the answer is “Yes”. You. For this reason, according to the present embodiment, it is possible to output a response such as “yes” or “no” by speech synthesis in a manner such as is seen when interacting with a person.

FIG. 10 is a diagram for explaining the output of the answers (3) and (4) to the question asked by the user, and is specific to the question "What is Asuno Tenki?" This is an example of a case where the answer is output by the speech synthesizer 10.
In the example of this figure, in response to the question, the answer of (4) of "um" is output before the answer of (3) of "Hare-de" as a specific answer. Since it is necessary to access various databases and external servers for specific answers, it may take time to create the answers. If the silent state continues in the speech synthesizer 10 until a specific answer is created, the user may feel uneasy. However, in the present embodiment, before the answer of (3), Since the answer of (4) is output for the time being, such kind of uneasiness is not given to the user.
Also, in a dialogue between people, it is a common act to give an answer as described in (4) before giving specific contents. In the present embodiment, since the response is synthesized by speech by imitating such an action, in that sense, the user can be given an impression as if he or she is interacting with a person.

In the figure, the timing at which the question was issued and the timing of the ending “ha” of the question are denoted by (d), the timing at which the beginning of the response to “er” is issued by (e), and (F) is the timing at which the beginning of the answer is issued. In this example, it is preferable that the period T3 from when the question is issued to when the answer of “er” is output is about the same as the answer in the above (1).
When it takes time to create a specific answer such as “sore”, as (4), the same answer or another answer may be repeated to lengthen the interval.

<Applications / Modifications>
The present invention is not limited to the above-described embodiment, and for example, various applications and modifications as described below are possible. In addition, one or a plurality of arbitrarily selected aspects of the application and modification described below can be appropriately combined.

<Voice input section>
In the embodiment, the voice input unit 102 is configured to input a user's voice (speech) with a microphone and convert it into a voice signal. However, the voice input unit 102 is not limited to this configuration, and may include a voice signal processed by another processing unit. Alternatively, a configuration may be adopted in which an audio signal supplied (or transferred) from another device is input. That is, the voice input unit 102 may have any configuration as long as the voice signal is input in some form.

<Sound waveform data>
In the embodiment, the answer creating unit 110 outputs a sound sequence in which a pitch is assigned to each sound as an answer to a question. However, the answer creating unit 110 outputs, for example, sound waveform data in a wav format of the answer. It is good also as composition.
In addition, since the sound waveform data does not assign a pitch to each sound as in the above-described sound sequence, for example, the sound control unit 109 specifies the pitch at the end of a simple reproduction, Configuration in which sound waveform data is output (reproduced) after performing pitch conversion (pitch conversion) such as filtering so that the specified pitch has a predetermined relationship with the pitch indicated by the pitch data. It is good.
The pitch may be converted by a so-called key control technique that shifts the pitch without changing the speech speed, which is well known in karaoke machines.

<End and beginning of answers etc.>
In the embodiment, the configuration is such that the pitch of the ending of the answer is controlled in accordance with the pitch of the ending of the question.However, depending on the language, dialect, wording, etc., parts other than the ending of the answer, for example, the beginning of the word is characteristic. In some cases. In such cases, the person who asked the question unconsciously compares the pitch at the end of the question with the pitch at the characteristic beginning of the answer when the question is answered. To judge the impression of the answer. Therefore, in this case, the pitch at the beginning of the answer may be controlled in accordance with the pitch at the end of the question. According to this configuration, when the beginning of the answer is characteristic, it is possible to give a psychological impression to the user who receives the answer. Specifically, see the drawings on the right side of FIGS. 6B, 7B, and 8B.

  The same applies to questions, not limited to endings, but may be determined by the beginning of a word. The question and answer are not limited to the beginning and end of the word, but may be determined by the average pitch or the pitch of the most pronounced part. For this reason, it can be said that the first section of the question and the second section of the answer are not necessarily limited to the beginning or end of the question.

<Relationship of pitch>
In the above-described embodiment, the speech synthesis is controlled so that the pitch of the answer ending or the like is 5 degrees lower than the ending of the question or the like. It may be a configuration. For example, as described above, the angle may be perfect 8 degrees, perfect 5 degrees, perfect 4 degrees, long / short 3 degrees, or long / short 6 degrees.
Further, even if the relationship is not the consonant pitch, there may be a case where the existence of a pitch relationship that gives a good (or bad) impression is recognized empirically, so that the pitch of the answer is controlled based on the pitch relationship. good. However, even in this case, if the pitch between the two pitches, such as the ending of the question and the ending of the answer, is too far apart, the answer to the question tends to be unnatural. It is desirable that the pitch of the answer be in the range of one octave up and down.

<Pitch shift of answer>
By the way, in a configuration in which the pitch such as the ending of the answer specified in the voice sequence or the like is controlled so as to have a predetermined relationship with the pitch such as the ending of the question, specifically, for example, as in the embodiment, In the configuration in which the pitch is changed by 5 degrees or 6 degrees below, if the pitch to be changed is too low, the answer may be synthesized with an unnatural bass sound. Therefore, next, application examples (No. 1 and No. 2) for avoiding such a case will be described.

FIG. 11 is a diagram showing a main part of the processing in the application example (Part 1). Note that the main part of the process referred to here is a process executed in “determination of answer pitch” in step Sa18 in FIG. That is, in the application example (No. 1), the process shown in FIG. 11 is executed in step Sa18 shown in FIG. 4, and the details are as follows.
First, the voice control unit 109 obtains a pitch which is 5 degrees lower than the pitch indicated by the pitch data from the pitch analysis unit 106, for example, if the answer is (1), and temporarily determines the pitch. If the answer is (2), the pitch is determined to be 6 degrees lower and temporarily determined (step Sb171).
Next, the voice control unit 109 determines whether or not the temporarily determined pitch is lower than a predetermined threshold pitch (step Sb172). Note that the threshold pitch is set to a pitch corresponding to a lower limit frequency when performing voice synthesis, or a pitch that gives an unnatural feeling if lower than this.

If the tentatively determined pitch, that is, the pitch 5 degrees or 6 degrees below the pitch at the end of the question is lower than the threshold pitch (if the determination result in step Sb172 is “Yes”), the voice control is performed. The unit 109 shifts the provisionally determined pitch to a pitch one octave higher (step Sb173).
On the other hand, if the obtained pitch is equal to or larger than the threshold pitch (if the determination result in step Sb172 is “No”), the processing in step Sb173 is skipped.
Then, the voice control unit 109 finally determines the pitch of the ending which is the target when shifting the answer to the following pitch (step Sb174). That is, if the tentatively determined pitch is lower than the threshold pitch, the voice control unit 109 changes the tentatively determined pitch to one octave higher and sets the tentatively determined pitch to the threshold pitch or more. If so, the target pitches are finally determined without changing the provisionally determined pitches.
After step Sb174, the processing procedure shifts to step Sa20 in FIG. 4, and the voice control unit 109 shifts the pitch of the answer to the finally determined pitch as the control content, and outputs the answer. The timing is determined to be the timing determined in step Sa19. Thereby, the voice synthesizing unit 112 synthesizes and outputs the voice of the voice sequence with the determined control content.

According to this application example (Part 1), if the pitch to be changed is lower than the threshold pitch, the pitch is shifted so that the pitch is one octave higher than the pitch. And that the answer is synthesized by speech.
Here, the pitch of the answer ending is shifted to the pitch one octave higher, but may be shifted to the pitch one octave lower. In detail, because the pitch of the ending of the question or the like issued by the user is high, if the pitch 5 degrees below the pitch is too high, the answer will be synthesized as an unnatural treble. . To avoid this, if the pitch (provisionally determined pitch) having a relationship of 5 degrees below the pitch indicated by the pitch data is higher than the threshold pitch, the pitch of the answer ending or the like is changed. The pitch may be shifted to a pitch one octave lower than the tentatively determined pitch.

  Further, when synthesizing voice, it may be possible to output a voice of a virtual character in which sex, age, etc. (child / adult) are determined. When a female or child character is specified as in this case, if the pitch is uniformly lowered to 5th or 6th below the ending of the question, the answer will be in a low tone that is not suitable for the character. May be synthesized so that the pitch is shifted by one octave.

  FIG. 12 is a diagram showing a main part of a process in such an application example (No. 2), and shows a process executed in “decision of pitch of answer” in step Sa18 in FIG. Explaining mainly the points different from FIG. 11, in step Sb171, the voice control unit 109 has a relationship of 5 degrees lower and 6 degrees lower than the pitch indicated by the pitch data from the pitch analysis unit 106. After the pitch is obtained and tentatively determined, it is determined whether a woman or a child is specified as an attribute defining the character (step Sc172).

If a female or child is specified as the attribute (if the determination result in step Sc172 is “Yes”), the voice control unit 109 shifts the temporarily determined pitch to a pitch one octave higher (step Sb173) On the other hand, if a woman or a child is not specified as the attribute, for example, if a man or an adult is specified (if the determination result in step Sc172 is “No”), the processing in step Sb173 is skipped. Is done. The subsequent steps are the same as in the application example (part 1).
According to this application example (2), if the answer is set to be made by the voice of a woman or a child, the pitch is shifted so that the pitch is one octave higher than the temporarily determined pitch. Can be avoided while maintaining the pitch relationship of 回答.
In this example, if a female or child is specified as an attribute, the pitch shifts to one octave higher. However, if an adult male is specified as an attribute, for example, a character corresponding to the attribute is not matched. The pitch may be shifted to one octave lower in order to avoid that the answer is synthesized with a treble that is suitable.

<Dissonance pitch>
In the above-described embodiment, the speech synthesis is controlled so that the pitch of the answer ending or the like has a consonant pitch relationship with the question ending or the like. May be controlled. It should be noted that if answers are synthesized at pitches that have a dissonant pitch relationship, the user who asked the question will feel unnatural, bad, or stupid, and a smooth dialogue will not be established. There are concerns, but some believe that such feelings can be good for stress relief.
In view of this, as an operation mode, a mode in which an answer such as a good impression is desired (first mode) and a mode in which an answer such as a bad impression is desired (second mode) are prepared, and speech synthesis is controlled in accordance with one of the modes. It is good also as a structure which performs.

  FIG. 13 is a diagram showing a main part of a process in such an application example (part 3), and shows a process executed in “decision of pitch of answer” in step Sa18 in FIG. Explaining mainly the points different from FIG. 11, the voice control unit 109 determines whether the first mode is set as the operation mode (step Sd172).

If the first mode is set as the operation mode (if the determination result in step Sd172 is “Yes”), the voice control unit 109 changes the pitch of the answer, for example, to the ending of the question, for example, to the pitch of the ending of the question, for example. On the other hand, the pitch is determined so as to have a consonant pitch (step Sd173A). On the other hand, if the second mode is set as the operation mode (if the determination result in step Sd172 is “No”), the voice control unit 109 sets the pitch at the end of the answer to the pitch at the end of the question. The pitch is determined so as to have a dissonant pitch relationship (step Sd173B). Subsequent steps are the same as in the application example (part 1) and the application example (part 2).
Therefore, according to this application example (part 3), if the first mode is set, the answer is voice-synthesized at a pitch having a consonant pitch with the questioned pitch, while the second mode is set. Is set, the answer is voice-synthesized at a pitch that has a dissonant pitch with the pitch of the question, so that the user can appropriately use the operation mode.

  Although the application example (part 1), the application example (part 2), and the application example (part 3) have been described using the example in which the audio sequence is used as in the first embodiment, the application example uses audio waveform data. Of course you can.

<Others>
In the embodiment, the language analyzer 108, the language database 122, and the answer database 124, which are configured to acquire the answer to the question, are provided on the side of the speech synthesizer 10, but the processing load is heavy in the terminal device and the like. In consideration of such a point that the storage capacity is limited, the configuration may be provided on the external server side. That is, in the speech synthesizer 10, the answer creation unit 110 only needs to have a configuration that obtains an answer to the question in some form and outputs data that defines the voice of the answer. It does not matter whether it is created on the side or on the side of a configuration other than the speech synthesizer 10 (for example, an external server).
The information acquisition unit 126 is unnecessary in the speech synthesizer 10 if the answer to the question can be created without accessing an external server or the like.

102: voice input unit, 104: utterance section detection unit, 106: pitch analysis unit, 108: language analysis unit, 109: voice control unit, 110: answer creation unit, 112: voice synthesis unit.

Claims (5)

An audio input unit for inputting an audio signal of the question,
An acquisition unit for acquiring an answer to the question,
A language analysis unit for analyzing the semantic content of the answer,
After finished the input speech signal of the questions is, a sound control section for controlling shorter than between the until the output audio signal of positive of the answers, to the output of the audio signal of the answer negative,
A voice control device comprising: The voice control unit,
Control is performed in the range from 0.5 seconds to 2.0 seconds until the voice signal of the affirmative answer is output.
The voice control device according to claim 1. The voice control unit,
When outputting the voice signal of the negative answer, the volume of the negative voice signal is set lower than the volume of the voice signal of the positive answer.
The voice control device according to claim 1. Input the audio signal of the question,
Get an answer to the question,
Analyze the meaning of the answer,
After finished the input speech signal of the question is between to the output of the audio signal of positive of the answers, and controlling shorter than until the output audio signal of the answer negative Voice control method. Computer
A voice input section for inputting the voice signal of the question,
An obtaining unit for obtaining an answer to the question,
A language analysis unit for analyzing the meaning of the answer, and
Voice control unit for controlling shorter than a period from finishing the input speech signal of the question is between to the output of the audio signal of the answer affirmative, to the output of the audio signal of the answer negative,
A program characterized by functioning as a program.

Images