JP2001117581A - Feeling recognition device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a feeling recognition device which can perform feeling recognition taking the degree of a feeling into account with a simple constitution. SOLUTION: A voice recognition part 12 specified a recognized character string by recognizing an input voice collected by a microphone 10. A kind judging part 16 judges the kind (affirmative or negative) of a feeling roughly according to the contents of the recognized character string. A degree decision part 18 detects the repetition of a vocabulary, a specific exclamation, and a specific adverb included in the input voice according to the contents of the recognized character string and judges the degree (high or low) of the feeling according to the detection result. A feeling recognition part 26 combines the judged results of the rough kind of the feeling and the degree of the feeling together to judge the detailed kind of the feeling.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an emotion recognition device for recognizing an emotion based on an input voice.

[0002]

2. Description of the Related Art Conventionally, an emotion recognition device for recognizing an emotion based on an input voice has been known. Such an emotion recognition device performs a voice recognition process to specify a character string corresponding to a voice, and performs emotion recognition based on the content of the character string, or performs voice strength, pitch, In addition, there are a method in which the degree of emotion (inflection) is detected by extracting a feature amount such as an utterance interval to perform emotion recognition, and a method in which both are combined to perform emotion recognition. In these emotion recognition devices, one emotion is identified from seven or eight kinds of emotions such as joy, anger, sadness, enjoyment, and surprise based on changes in the content of a character string and each feature amount of voice. Then, this is output as an emotion recognition result.

[0003]

In the above-described conventional emotion recognition apparatus, in which emotion recognition is performed based on the contents of a character string, vocabulary information and a character string prepared in advance as vocabularies indicating emotions are used. Although it is possible to identify the type of emotion by judging whether or not the content of the emotion matches, the problem is that the degree of emotion cannot be identified. Further, when emotion recognition is performed based on the feature amount of speech, it is relatively easy to determine the degree of emotion, but it is difficult to specify the type of emotion. In addition, in the case of using both the contents of a character string and the feature amount of a voice as information, a configuration for recognizing the content of a character string and a configuration for extracting a feature amount of a voice are required. There is a problem that the configuration of the apparatus is complicated.

SUMMARY OF THE INVENTION The present invention has been made in view of the above circumstances, and an object of the present invention is to provide an emotion recognition device capable of performing emotion recognition with a simple structure in consideration of the degree of emotion. is there.

[0005]

In order to solve the above-mentioned problems, an emotion recognition apparatus according to the present invention performs a voice recognition process on an input voice by a voice recognition unit to specify a recognition character string, and performs the recognition. The first emotion determination unit determines a rough type of emotion based on the content of the character string, and the degree determination unit determines the degree of emotion based on the content of the recognized character string. The detailed emotion type is determined by the second emotion determination unit based on the general emotion type determined by the determination unit and the degree of emotion determined by the degree determination unit. As described above, since the degree of emotion is determined based on the content of the recognized character string, the voice signal is compared with the conventional case where the degree of emotion is determined based on the loudness or pitch of the voice. There is no need for a complicated configuration for analyzing and extracting each feature,
With a simple configuration, emotion recognition can be performed in consideration of the degree of emotion.

It is desirable that the degree determining means determines the degree of emotion based on at least one of vocabulary duplication, exclamation and adverb contained in the input speech. Considering the conversation between humans, if the degree of emotion is high, in the voice, for example, duplication of vocabulary such as "no useless", exclamation such as "wow" and adverb such as "great" appear, The degree of emotion can be easily determined by detecting these elements (duplicate vocabulary, exclamation, adverb) contained in the voice.

[0007] Further, the general types of emotions determined by the first emotion determining means are at least two types, positive and negative, and the degree of emotion determined by the degree determining means is at least: High or low, and the second emotion determination means is a positive and high, a positive and low, a negative and high, a negative obtained by combining the general type of emotion and the degree of emotion. It is desirable to determine detailed emotions by selecting one of at least four types of target and low. For example, assuming a case where various actions are determined based on a user's emotion in various systems, it is often sufficient to recognize about four types of detailed emotions as described above. By applying the emotion recognition device, it is possible to easily realize various systems that perform various operations by performing emotion recognition in consideration of the degree of the user's emotion.

[0008]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, an emotion recognition apparatus according to an embodiment of the present invention will be described with reference to the drawings. FIG. 1 is a diagram illustrating a configuration of an emotion recognition device according to an embodiment. The emotion recognition device 1 shown in FIG.
0, a voice recognition unit 12, a voice recognition dictionary storage unit 14, a type determination unit 16, a degree determination unit 18, and an emotion recognition unit 26.

[0009] The microphone 10 collects the voice uttered by the user and converts it into a voice signal. The voice recognition unit 12
The voice signal output from the microphone 10 is analyzed to perform a predetermined voice recognition process, and a recognition character string corresponding to the voice uttered by the user (input voice) is specified. The voice recognition dictionary storage unit 14 stores a signal waveform corresponding to a standard voice as a voice recognition dictionary.

[0010] The type determination section 16 determines the approximate type of emotion based on the recognition character string output from the voice recognition section 12. Here, in the present embodiment, as the general types of emotions determined by the type determination unit 16, two types of positive (Positive) and negative (Negative) are considered.
Hereinafter, a positive emotion is referred to as “emotion P” and a negative emotion is referred to as “emotion N”. The type determination unit 16
By roughly determining whether the vocabulary corresponding to the recognized character string belongs to the emotion P or the emotion N described above, the general type of emotion is determined.

FIG. 2 is a diagram for explaining the relationship between various vocabularies and general types of emotions. As shown in FIG. 2, the vocabulary belonging to the emotion P (positive emotion) is
"Like", "Look good", "Cool", "Look fun", "Love", "Best", etc. can be considered. Vocabulary belonging to emotion N (negative emotion) is
"First,""Cool,""Sorry,"
"Dislike", "lowest", etc. are considered. Note that the vocabulary belonging to the emotion P and the vocabulary belonging to the emotion N are not limited to the above example, and various other vocabularies can be considered.
The type determination unit 16 stores the relationship between the various vocabularies as shown in FIG. 2 and the general types of emotions as a data table, and refers to this data table to determine the vocabulary corresponding to the recognized character string. The type of the emotion is roughly determined by examining which one of the emotion P and the emotion N belongs.

The degree judging section 18 checks whether or not the input speech contains at least one of vocabulary duplication, a predetermined exclamation, and a predetermined adverb, based on the recognition character string output from the voice recognition section 12. This determines the degree of emotion.
Specifically, the degree determining unit 18 determines that the degree of emotion is high (Exciting) when the input voice includes at least one of the above-described vocabulary duplication, a predetermined exclamation, and a predetermined adverb. Otherwise, the degree of emotion is determined to be low (Depressing). Hereinafter, a high degree of emotion is referred to as "degree E", and a low degree of emotion is referred to as "degree D".

In order to detect each of vocabulary duplication, predetermined exclamation, and predetermined adverb contained in the input voice, the degree judging unit 18 includes a duplication detecting unit 20, an exclamation detecting unit 22, and an adverb detecting unit. 24. Duplication detector 20
Detects, based on the recognition character string output from the voice recognition unit 12, when the input voice has a duplicate portion of the same vocabulary. Specifically, for example, when a recognition character string corresponding to an input voice such as “like” or “no useless” is output, the duplication detection unit 20 regards this as an overlapping part of the same vocabulary. To detect. The exclamation point detection unit 22
Based on the recognition character string output from the voice recognition unit 12,
A predetermined exclamation included in the input voice is detected. FIG.
FIG. 4 is a diagram illustrating an example of an exclamation detected by the exclamation detection unit 22. The exclamation word detecting unit 22 determines a predetermined exclamation word (such as “Wow ぁ”, “Oo”, or “Hie”) as shown in FIG.
A data table is stored, and when any of the exclamation words stored in the data table is included in the input voice, this is detected. The adverb detection unit 24 detects a predetermined adverb included in the input speech based on the recognition character string output from the speech recognition unit 12. FIG. 4 is a diagram illustrating an example of an adverb detected by the adverb detecting unit 24. The adverb detecting unit 24 stores a data table related to a predetermined adverb ("very much", "very", etc.) as shown in FIG. 4, and any of the adverbs stored in this data table is included in the input voice. Detect if it is included.

The emotion recognition unit 26 determines detailed emotions based on the general type of emotion determined by the type determination unit 16 and the degree of emotion determined by the degree determination unit 18. Specifically, the emotion recognizing unit 26 outputs “emotion P” or “emotion N” that is output as a determination result regarding the general type of emotion and “degree E” or “degree E” which is output as a determination result regarding the degree of emotion. Based on the "degree D", detailed emotions are combined with "emotional P" and "degree E" (hereinafter, referred to as "emotional P / E"), and the combination of "emotional P" and "degree D" ( Hereinafter, this is referred to as “emotional P / D”, a combination of “emotional N” and “degree D” (hereinafter, referred to as “emotional N / D”), and a combination of “emotional N” and “degree E”. Combinations (hereinafter referred to as “emotional N / E”) are classified into four types, and detailed emotions are determined by determining which of the combinations corresponds. The detailed emotion determined by the emotion recognition unit 26 is output as an emotion recognition result by the emotion recognition device 1.

FIG. 5 is a diagram showing an example of the emotion recognition result determined by the emotion recognition section 26. As shown in FIG. In the table shown in FIG. 5, various criteria such as “good / bad” and “good / unsatisfactory” are shown in the first column from the left, and the relationship between the input voice and the emotion recognition result with respect to these criteria is shown. An example is shown. For example, when the input voice for the judgment criterion “good / bad” is “like”, it is determined from the vocabulary of “like” that the general emotion type belongs to “emotion P”, and the same vocabulary is used. Are determined to belong to the “degree E” because of the overlap, so that “emotion P / E” is obtained as a detailed emotion determination result. Similarly, if the input voice for the criterion “like / dislike” is “dislike”, it is determined from the vocabulary of “dislike” that the general type of emotion belongs to “emotion N”, Since neither the exclamation verb nor the adverb is included in the input voice, the degree of the emotion is determined to belong to “degree D”.
D "is obtained.

The speech recognition unit 12 and the speech recognition dictionary storage unit 14 described above serve as speech recognition means, the type determination unit 16 serves as first emotion determination means, the degree determination unit 18 serves as degree determination means, and the emotion recognition unit 26 includes Each corresponds to the second emotion determination means.

The emotion recognition apparatus 1 of the present embodiment has the above-described configuration, and the operation will be described next. FIG.
5 is a flowchart showing an operation procedure of the emotion recognition device 1 of the present embodiment.

The voice recognition unit 12 always determines whether or not a voice has been input to the microphone 10 (step 100). If a voice has been input, the voice is stored in the voice recognition dictionary storage unit 14. A speech recognition process is performed using the speech recognition dictionary thus specified, and a recognition character string corresponding to the input speech is specified (step 101).

Next, the type judging section 16
Based on the recognition character string output from the above, a rough emotion type is determined using a data table showing the relationship between the various vocabularies and the rough emotion type described above (step 1).
02). For example, if the recognized character string indicates a vocabulary belonging to the emotion P (positive emotion) such as “like” or “like”, the general emotion type is “emotion P ”And the recognized character strings are“ No ”and“ Dislike ”.
If the word indicates a vocabulary belonging to the emotion N (negative emotion), the general emotion type is determined to be “emotion N”.

Further, the degree judging section 18 comprises the speech recognition section 12
Then, the degree of emotion is determined based on the recognition character string output from (step 103). For example, the degree determining unit 18
Means that the recognition string contains vocabulary duplication such as "like",
If at least one of an exclamation word such as "Wow" and an adverb such as "Wow" is included, the degree of emotion is determined as "Degree E" (the degree of emotion is high), and the vocabulary is determined. If neither the duplication, the exclamation, nor the adverb is included, the degree of emotion is determined to be “degree D” (the degree of emotion is low). Note that the processing shown in step 102 described above and step 1
The order of performing the processes may be changed from the process indicated by 03, or the processes may be performed simultaneously.

[0021] After roughly determining the type of emotion and determining the degree of emotion and outputting each determination result, the emotion recognizing unit 26 determines detailed emotion based on these determination results. (Step 104). Specifically, as described above, the rough emotion type determination result is output as “emotion P” or “emotion N”, and the determination result of the degree of emotion is “degree E” or “degree D”. Is output, the emotion recognition unit 26 determines that the detailed emotion is “emotional P / E”, “emotional P / D”, “emotional N / E”, “emotional N / D” from the combination of these determination results. Is determined. When the detailed emotion determination operation is completed, step 1
Returning to 00, the operation after the determination of the presence or absence of the voice input is repeated.

Next, an example in which the emotion recognition device 1 of the present embodiment is applied to a vehicle-mounted system will be described. FIG.
FIG. 1 is a diagram illustrating a configuration of a vehicle-mounted system including an emotion recognition device 1 according to an embodiment. The in-vehicle system 100 illustrated in FIG. 7 includes an emotion recognition device 1 and a voice processing device 2 that determines and outputs various operation instructions in an interactive manner in response to a voice uttered by a user by performing voice recognition. A navigation device 3 for detecting the position of the vehicle and displaying a map around the position of the vehicle, searching for a route to the destination selected by the user, and guiding the route, a radio tuner, a cassette tape deck, and a CD. An audio device 4 containing a player or the like, a display device 5 for displaying various images output from the navigation device 3 or the like, and a guidance sound or the like output from the navigation device 2 or an audio sound or the like output from the audio device 4 It is configured to include a speaker 6 that outputs various sounds.

Next, a description will be given of an example in which the voice processing device 2 performs a dialogue with a user in response to the emotion recognition result output from the emotion recognition device 1 in the above-described on-vehicle system 100.

As a specific example, the voice processing device 2 determines the destination of the drive based on the contents of the dialogue with the user,
An example in which an operation instruction is output to the navigation device 3 or the audio device 4 by deciding music to be played will be described. In the description of the dialogue example, the user is referred to as “U”.
(User), and the audio processing device 2 is indicated by “A” (Agent).

(Specific Example 1: Determining the Destination of the Drive) U: "I want to go somewhere to play." A: "Let's go outside the prefecture?" U: "Oh, I like it ." (A determination result of “emotion P / E” is obtained. From the determination result, the voice processing device 2
Judges that the user strongly desires to go out, and determines a specific proposal. ) A: "? Is the per Miyagi how" U:. ". Damedame, will not may snow" (judgment result of "emotion N / E" is obtained voice processing apparatus 2 from the result of this determination is, Since the user's emotions are extremely negative, we judge that it is necessary to apologize and present another destination candidate, and learn to drive NG in the late autumn in the direction of Tohoku.) A: "I'm sorry, I don't care. A: "Well then, let's go to Tokyo." U: "Yeah, I like ." (A judgment result of "Emotion P / D" is obtained. From this judgment result, the speech processing device 2 Judge that the user's emotion is positive and decide the destination of the drive to Tokyo.) A: "Then we will guide you to Tokyo." (The voice processing device 2 operates the navigation device 3 Output instructions and go to Tokyo To carry out the route guidance) (Specific Example 2:. Determination of the music to be played) U: "is I want to hear music." A: "Do you want to apologize in jazz" U: ". Like" ( "Emotion A determination result of “P / D” is obtained. From this determination result, the voice processing device 2 determines that the user's emotion is positive and the music genre is jazz.) A: “○○○ (singer) ? name) Why even "U:!" is not that a good hobby or really I ○○○
It's cool ! (A determination result of “emotion P / E” is obtained. From this determination result, the voice processing device 2 determines that the user's emotion with respect to his / her proposal is very positive, and determines music to be played. At the same time, learn the user's preferences (I love the singer's OO).) A: "Teehe (Shining voice)! Then, I'll play the OO song." The device 2 outputs an operation instruction to the audio device 4 to play the tune of the singer XX.) As described in the specific examples 1 and 2, the emotion is output from the emotion recognition device 1 of the present embodiment. By using the emotion recognition result, the voice processing device 2 performs a dialog reflecting the user's emotion, and instructs each device included in the vehicle-mounted system 100 to perform various operation instructions according to the user's intention. Can be given. Therefore, a more human-friendly system can be realized.

As described above, the emotion recognition apparatus 1 of the present embodiment
Is determined by the type determination unit 16 on the basis of the content of the recognized character string specified by the speech recognition unit 12, and the degree of emotion (positive or negative) is determined by the degree determination unit 18. (High or low), and the emotion recognition unit 26 determines a detailed emotion type based on each determination result of the general emotion type and emotion degree. Therefore, a complicated configuration for analyzing a voice signal and extracting each feature amount is unnecessary compared with a case where a degree of emotion is determined based on a loudness or a pitch of a voice as in the related art. With the configuration, emotion recognition can be performed in consideration of the degree of emotion. In particular, vocabulary duplication, exclamation and adverb included in the input speech are used as elements for determining the degree of emotion, and these elements (duplication of vocabulary, exclamation and adverb) are used in conventional speech recognition. Since the detection can be easily performed by using the technique, the degree of the emotion can be easily determined.

The present invention is not limited to the above embodiment, and various modifications can be made within the scope of the present invention. For example, in the embodiment described above,
The degree determining unit 18 determines the degree of emotion by examining whether or not the input voice includes at least one of vocabulary duplication, a predetermined exclamation, and a predetermined adverb. (Duplicate, predetermined exclamation, predetermined adverb) does not necessarily need to be examined, and any one or two of these elements may be examined to determine the degree of emotion.

Further, in the above-described embodiment, the on-vehicle system 100 to which the emotion recognition device 1 of the present invention is applied has been described. However, the application of the present invention is not limited to this, and is applied to various systems. can do. For example, when the emotion recognition device of the present invention is applied to a robot, a human-like behavior that takes into account the type and degree of emotion in accordance with four types of emotion obtained as detailed types of emotions is required. A robot that can do this can be easily realized.

[0029]

As described above, according to the present invention, the approximate emotion type (positive or negative) and the emotion type are determined based on the contents of the recognized character string specified in response to the input voice. The degree (high or low) is determined, and a detailed emotion type is determined based on these determination results.
Therefore, a complicated configuration for analyzing the audio signal and extracting each feature amount is unnecessary compared with the conventional case where the degree of emotion is determined based on the volume or pitch of the voice, and the like.
With a simple configuration, emotion recognition can be performed in consideration of the degree of emotion. In particular, since the vocabulary duplication, the exclamation, and the adverb included in the input speech are used as elements for determining the degree of emotion, the degree of emotion can be easily determined.

[Brief description of the drawings]

FIG. 1 is a diagram illustrating a configuration of an emotion recognition device according to an embodiment.

FIG. 2 is a diagram illustrating a relationship between various vocabularies and types of general emotions.

FIG. 3 is a diagram illustrating an example of an exclamation detected by an exclamation detection unit.

FIG. 4 is a diagram illustrating an example of an adverb detected by an adverb detecting unit.

FIG. 5 is a diagram illustrating an example of an emotion recognition result determined by an emotion recognition unit.

FIG. 6 is a flowchart showing an operation procedure of the emotion recognition device.

FIG. 7 is a diagram showing a configuration of a vehicle-mounted system including the emotion recognition device of the present embodiment.

[Explanation of symbols]

 Reference Signs List 1 emotion recognition device 10 microphone 12 voice recognition unit 14 voice recognition dictionary storage unit 16 type determination unit 18 degree determination unit 20 duplication detection unit 22 exclamation detection unit 24 adverb detection unit 26 emotion recognition unit

Claims (3)

[Claims] 1. A speech recognition means for performing a speech recognition process on an input speech to specify a recognition character string corresponding to the input speech, and a general emotion type based on the content of the recognition character string. First emotion determining means for determining; degree determining means for determining the degree of emotion based on the content of the recognized character string; and the general emotion type determined by the first emotion determining means; A second emotion determination unit that determines a detailed type of emotion based on the degree of the emotion determined by the degree determination unit. 2. The emotion according to claim 1, wherein the degree determining means determines the degree of the emotion based on at least one of a vocabulary overlap, an exclamation, and an adverb included in the input speech. Recognition device. 3. The general emotion type determined by the first emotion determination unit according to claim 1 or 2, wherein the general type of emotion is at least two of positive or negative.
The degree of the emotion determined by the degree determination means is at least two types, high or low, and the detailed emotion type determined by the second emotion determination means is at least: Emotion recognition including four types of positive and high, positive and low, negative and high, negative and low obtained by combining the general type of emotion and the degree of the emotion apparatus.