JP2003263308A - Screen control device and method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a screen control device making it possible to provide instructions on processes for a display screen and a processing amount by means of voice and a more intuitive and easier method. <P>SOLUTION: Input voice has its features extracted as it is analyzed by the acoustic feature extracting part 11 of the screen control device 10, with the features being inputted to the neural network of a phoneme distinguishing part 12 for each one frame. Phonemes are distinguished by the phoneme distinguishing part 12 in real time and results for five frames are output as a line of phoneme distinction patterns G. A phoneme continuation detecting part 13 determines the average value of output levels among five frames for each phoneme according to the line of phoneme distinction patterns G, and if the output level of the average is greater than a predetermined value H, a determination is made that the phoneme was input in continuation. Only for the period during which the signal of the phoneme continues, a screen control part 14 performs a predetermined screen operation that is allocated to that phoneme. <P>COPYRIGHT: (C)2003,JPO

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a screen control device for instructing an operation for a display image on a screen by voice, and further for instructing the amount of the operation by the duration of a predetermined sound of the voice. The present invention relates to a screen control device capable of performing.

[0002]

2. Description of the Related Art If it is possible to directly instruct the operation of various equipment by voice, the operation of the equipment can be performed easily and efficiently, and the human interface is remarkably improved.
Then, with the recent development of voice recognition technology, a method of actually controlling various devices by voice has been proposed.
For example, on the screen of a workstation or a personal computer, a method of moving, enlarging, reducing, or rotating the various displayed objects by voice input has been proposed.

In this method, for example, when an operator utters a word indicating a predetermined process to a voice input means such as a microphone externally attached to a workstation or the like, the voice is recognized and predetermined. The process corresponding to the recognized word is determined by referring to the correspondence between the word and the process, and the screen is controlled to execute the process. That is, the input voice is recognized as one command, and the same processing as command input from a normal keyboard or the like is performed to control the screen.

[0004]

However, the above-mentioned instruction of screen control by voice so far is performed by a command input (voice input) once by a predetermined amount of movement / movement.
Enlargement / reduction / rotation is performed, and it is impossible to move it by an arbitrary distance, enlarge / reduce it at an arbitrary magnification, or rotate it at an arbitrary angle.
In that way, if you try to perform the above process by any amount,
It was necessary to separately specify a numerical value from the keyboard, to separately utter a numerical value for recognition, or to separately input instructions for starting and ending the processing. Such processing is an operation comparable to a normal screen operation using a keyboard or the like, and it is hard to say that it is a device with a good human interface that takes full advantage of the convenience of command input by voice.

Therefore, it is an object of the present invention to provide a screen control capable of instructing a display device for displaying an arbitrary image a process for a display screen and its processing amount by voice in an intuitive and easy method. To provide a device. Another object of the present invention is to provide a display device,
It is to provide a screen control method for instructing both the type and the amount of processing by voice.

[0006]

The inventor of the present application has studied the use of a utterance period that can be easily adjusted in an analog manner in order to specify the amount of processing together with the processing by command input by voice. Especially when controlling the screen,
Since it is possible to observe the processing result in real time, it is possible to directly and intuitively specify the processing that brings the screen to a desired result if the utterance is continued / finished while watching the screen.

According to the first aspect of the present invention, the image displayed on the screen is enlarged or reduced in accordance with the content of the word of the voice registered in advance, and is enlarged or reduced according to the duration of the ending of the voice. A screen control device for enlarging or reducing, a voice recognition means for recognizing the content of a word of an input voice and determining whether to enlarge or reduce the image displayed on the screen, and the ending of the input voice. A duration detection means for detecting the duration and an image displayed on the screen,
There is provided a screen control device having screen control means for enlarging or reducing according to the detected duration of the ending.

According to the second aspect of the present invention, the image displayed on the screen is enlarged or reduced according to the content of the word of the voice registered in advance, according to the duration of the ending of the voice. A voice control step of recognizing the content of a word of an input voice to determine enlargement or reduction of an image displayed on the screen, and a ending of the input voice. A screen control method is provided, which includes a duration detection step of detecting the duration of time, and a screen control step of enlarging or reducing the image displayed on the screen according to the detected duration of the ending. It

[0009]

According to the present invention, when a voice is input to the controlled display device, the input voice is analyzed, feature extraction is performed, and the input sound is identified based on the feature. Then, it is determined whether or not the identified audio signal is an audio signal corresponding to a predetermined process for the display device, and if it is a corresponding audio signal, the predetermined process is executed for the display device. . On the other hand, when a predetermined sound sequence is detected from the input audio signal that specified the process, and when the predetermined process is started, the process is continued while the predetermined sound sequence continues. To do.

[0010]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS First Embodiment A first embodiment of the screen control device of the present invention will be described with reference to FIGS. FIG. 1 is a block diagram showing the configuration of the screen control device of the first embodiment. The screen control device 10 includes an acoustic feature extraction unit 11, a phoneme identification unit 12, and a phoneme continuation detection unit 1.
3 and a screen control unit 14. The configuration and operation of each unit will be described below.

The acoustic feature extraction unit 11 extracts an acoustic feature amount from the input voice signal and outputs it to the phoneme identification unit 12. Table 1 shows the conditions of the analysis performed by the acoustic feature extraction unit 11. Also, by this analysis, the phoneme identification unit 12 uses 5 frames (50 ms) of the output of the mel-scale 16-channel FFT as a feature amount.
Output to.

[0012]

[Table 1]

The phoneme identification unit 12 identifies phonemes based on the acoustic feature quantity input from the acoustic feature extraction unit 11,
It is a phoneme recognition unit that outputs the identification result to the phoneme continuation detection unit 13. The phoneme identification unit 12 is composed of a four-layer feedforward neural network as shown in FIG. This neural network has an input layer,
80 hidden layers, 1 hidden layer 2, and 80 hidden layers,
It has 100, 50, and 6 units. The phoneme identification unit 12 identifies the five vowels and the six phonemes of / sh / shown in Table 2 by this neural network.

[0014]

[Table 2]

In this neural network, it is necessary to perform learning in advance to construct a circuit network capable of identifying phonemes. In the present embodiment, by learning the 6 phonemes shown in Table 2 using 2620 words by one specific male speaker, a circuit network capable of sufficiently identifying the 6 phonemes is constructed. I was able to. The output of the phoneme identification unit 12 is the phoneme identification pattern sequence G = (g ₁ ,
g ₂ , ..., G _t ..., G _T ).
The element g _t of the phoneme identification pattern string G represents the output vector at time t, and the element number T represents the voice frame length. Also,
The number of vector dimensions of the output vector g _{t is} the number of output units of the neural network, which is 6 in this embodiment.

The phoneme continuation detection unit 13 detects continuation of the same phoneme based on the phoneme identification pattern string G which is the output of the phoneme identification unit 12, and during that continuation, a signal indicating that is detected is screen-controlled. It is output to the unit 14. Specifically, the output level of each unit of the output layer of the neural network of the phoneme identification unit 12 is observed, the time during which the average of the n frames maintains a constant level H is measured, and the time is the minimum duration time. When L _min or more continues, it is determined that the voice signal of the phoneme is continuously input. Then, while the continuation is detected, a signal indicating the continuation of the phoneme is output to the screen control unit 14.

The operation of the phoneme continuation detecting unit 13 will be specifically described with reference to FIG. FIG. 3 is a diagram for explaining the detection of phoneme continuation, in which one of the six-dimensional vector g _t corresponding to the output unit of the neural network of the phoneme identification unit 12 is used.
It is a figure which shows the output value for every 0 ms. Further, in this embodiment, the minimum duration L _min is 5 frames.

In the example shown in FIG. 3, the output of the phoneme / u / is started from the frame 4 and the level is increased between the frame 6 and the frame 30. Further, in the frame 31 and the frame 32, low-level output is observed. Therefore, if the average value of the output levels from the frame 5 to the frame 9 is larger than the predetermined value H, the phoneme / u /
Is detected, and a signal indicating that it is continuing is output. Further, when the frame 31 and the frame 32 are continuously connected to frames having a low output level, and the average value of the output levels becomes smaller than the predetermined value H, the signal indicating the continuation is immediately terminated. As a result, the phoneme /
The signal indicating the continuation of u / is output between the frame 9 and the frame 31.

When a signal indicating that a specific phoneme is continuously input is input from the phoneme continuation detection unit 13, the screen control unit 14 receives a predetermined signal corresponding to the phoneme during the input period. The process is performed on the screen of the display device. Table 3 shows the processing of the screen corresponding to the detected phonemes.

[0020]

[Table 3]

The process for this display screen will be described in detail with reference to FIG. FIG. 4 is a diagram for explaining the operation of the display screen, and shows a state in which the pointer 41, the first window 42, the second window 43, and the clock tool 44 are displayed on the display screen 40. The display screen 40 is a screen normally seen when the window system is started on a workstation or a personal computer.

When the phoneme continuation detector 13 inputs a signal indicating continuation of a phoneme of / a /, the screen controller 14 moves the pointer 41 to the right at a predetermined speed while the signal is being input. Similarly, when a signal indicating continuation of a phoneme of / i / is input, the pointer 41 is moved to the left while the signal is input, and when a signal indicating continuation of a phoneme of / u / is input, the pointer 41 is moved up. In each direction at a predetermined speed. When a signal indicating the continuation of the phoneme of / o / is input, the window in which the pointer 41 exists, that is, the first window 42 in the display screen 40 of FIG. 4, is gradually expanded at a predetermined speed, and / sh / When the signal indicating the continuation of the phoneme is input, the first window 42 in which the pointer 41 exists is also reduced at a predetermined speed.

The operation of the screen control device 10 having such a configuration will be specifically described. When a voice is input to the display device controlled by the screen control device 10, the acoustic feature extraction unit 11 analyzes and extracts the feature, and inputs the feature to the neural network of the phoneme identification unit 12 for each frame (10 ms). To be done. In the phoneme identification unit 12, the phoneme is identified in real time by the neural network, and the identification result for five frames is output as the phoneme identification pattern string G together with the identification results of the frames so far. Then, in the phoneme continuation detection unit 13,
From the phoneme identification pattern sequence G, the average value of the output levels for the respective 5 frames for each phoneme is obtained. When the calculated average output level is higher than the predetermined value H, it is determined that the phoneme is continuously input, and a signal to that effect is output to the screen control unit 14. The screen control unit 14 performs a predetermined screen operation for each phoneme based on the signal.

As described above, according to the screen control device of the first embodiment, it is possible to perform a predetermined process on the screen display only by uttering a predetermined single tone, and further, the amount of the process is continuous. Can be adjusted according to the length of utterance. Therefore, by continuing or stopping uttering while watching the screen, it is possible to perform a predetermined amount of processing on the screen. Further, the screen control device of the first embodiment recognizes simple phonemes and detects their continuity, so that the structure of the device is simple. In particular, even if a neural network is used as the phoneme identifying means as in this embodiment, it can be constructed by a relatively simple neural network.

Second Embodiment A second embodiment of the screen control device of the present invention will be described with reference to FIGS. FIG. 5 is a block diagram showing the configuration of the screen control device of the second embodiment. The screen control device 50 includes an acoustic feature extraction unit 51, a phoneme identification unit 52, a word identification unit 53, a screen control unit 54, and a voice input detection unit 55. The configuration and operation of each unit will be described below.

The configuration of the acoustic feature extraction unit 51 is basically the same as the configuration of the acoustic feature extraction unit 11 of the first embodiment described above, and the input voice signal is analyzed under the conditions shown in Table 1, The feature amount is extracted and output to the phoneme identification unit 52. However, in the second embodiment, the FFT is used as the feature amount.
It is assumed that 7 frames (70 ms) of output are output as the feature amount.

The phoneme identifying section 52 identifies phonemes based on the acoustic feature quantity input from the acoustic feature extracting section 51,
It is a phoneme recognition means that outputs the identification result to the phoneme continuation detection unit 53, and is composed of a four-layer feedforward neural network as in the first embodiment. However, the neural network forming the phoneme identification unit 52 of the second embodiment has 112 units, 1250 units, 100 units, and 25 units in the input layer, the hidden layer 1, the hidden layer 2, and the output layer, respectively. Then, the 25 phonemes shown in Table 4 are identified by this neural network.

[0028]

[Table 4]

Also, in the neural network of the phoneme identification unit 52, 2 voices are produced by one specific male speaker.
Learning is performed using 620 words to form a circuit network in which 25 phonemes shown in Table 4 can be identified. The phoneme identification unit 1
The output of 2 is the phoneme identification pattern sequence G = as in the first embodiment.
(G ₁ , g ₂ , ..., G _t ..., G _T ) are output. Since the number of dimensions of the output vector g _{t is} the number of output units of the neural network, it is 25 in the second embodiment.

The word identifying section 53 detects a predetermined word based on the phoneme identifying pattern string G output from the phoneme identifying section 52, and outputs a signal to that effect to the screen control section 54. Specifically, the output level of each unit of the output layer of the neural network of the phoneme identification unit 52 is observed, the time during which the average of the n frames maintains a constant level H is measured, and the time is the minimum duration time. When L _min or more continues, it is detected that the voice signal of the phoneme is input. Then, the sequence of phonemes that is sequentially detected is compared with the character string of the word to be identified, which is stored in advance in a storage unit (not shown) in the word identifying unit 53, and the sequence of phonemes matches any of the character strings. Then, it is determined that the word is input as a voice signal, and a signal to that effect is sent to the screen control unit 1.
Output to 4.

The voice input detector 55 outputs a voice detection signal to the screen controller 54 while the voice input is detected. That is, the energy level of the input voice signal is measured, the measurement result is compared with a predetermined threshold level, and the voice detection signal is detected when the level of the input voice signal is higher than the threshold value. To detect.

The screen control unit 54 performs a predetermined process on the display device for a predetermined period based on the word identification result input from the word identification unit 53 and the voice detection signal input from the voice input detection unit 55. To do. The screen control unit 54 determines a process to be executed according to the word input from the word identifying unit 53, and performs the process on the screen of the display device while the voice detection signal is input from the voice input detection unit 55. To execute. That is, the screen control unit 54, after a predetermined word is input, while the phoneme at the end of the word is continuously emitted,
Continue executing the screen operation determined by the word. Table 5 shows the processing of the screen corresponding to the identified word.

[0033]

[Table 5]

The operation of the screen control device 50 of the second embodiment will be specifically described with reference to FIGS. 6 and 7. FIG. 6 is a diagram for explaining phoneme detection, word identification, and word ending continuation. In the screen control device 50, the input voice signal is analyzed by the acoustic feature extraction unit 51, and when the output vector for each frame is obtained, the phoneme identification unit 52 identifies the phoneme in units of every seven frames. In the example shown in FIG. 6, the output level corresponding to the phoneme / u / is observed during 7 frames between the frame 5 and the frame 11, and the output level corresponding to the phoneme / e / is observed between the frame 12 and the frame 32. Output level is observed. Therefore, the phoneme identification unit 52 recognizes the phoneme / u / in the frame 11 and the phoneme / e / in the frame 18, and as a result, the word identification unit 53 recognizes the word “up” in the frame 18. On the other hand, the voice input detection unit 55 observes the energy level of the input voice signal, detects an energy level equal to or higher than a predetermined threshold value between the frame 5 and the frame 32, and detects the voice signal of the voice signal during this period. A signal indicating that there is an input is output to the screen control unit 54.

In the screen control section 54, the word identifying section 5
When the detection signal of the word "Ue" is input from 3, first,
A signal for executing the screen operation corresponding to the word "Ue" is output. Then, the signal from the voice input detection unit 55 is checked from the time when the detection signal of the word "Ue" is input, and it is checked whether or not the voice input continues to exist in the word. That is, it is checked whether or not the ending of the word "Ue" has been extended and entered. Then, the screen control unit 54 uses the voice input detection unit 55.
The signal for executing the screen operation continues to be output while the signal from is input following the signal for detecting the word. In the example of FIG. 6, the frame 19 to the frame 32
Until, the screen operation is continued as the period of ending words. In the frame 33, when the signal indicating that there is a voice input from the voice input detection unit 55 ends, the screen control unit 54
The transmission of the screen operation signal is terminated.

The screen operation will be described with reference to FIG. FIG. 7 is a diagram for explaining the operation of the display screen. The display screen 70 includes a pointer 71, a first window 72,
The state where the second window 73 and the clock tool 74 are displayed is shown. The screen control unit 54 moves the pointer 71 to the right at a predetermined speed when a signal indicating that the word “Migi” is identified is input from the word identifying unit 53. Similarly, when a signal indicating the word "hidari" is input, the pointer 71 is moved to the left, when a signal indicating the word "up" is input, the pointer 71 is moved upward, and a signal indicating the word "do" is input. The pointer 71 is moved downward at a predetermined speed.

When a signal indicating the identification of the word "Kakudai" is input, the window in which the pointer 71 exists,
That is, in the display screen 70 of FIG. 7, the first window 72 is gradually expanded at a predetermined speed, and when a signal indicating the identification of the word “shushusho” is input, the first window 72 in which the pointer 71 exists is also changed. Reduce at a predetermined speed. Move these pointers, expand windows /
The reduction process is continuously performed based on the signal from the voice input detection unit 55 while the ending of each word is extended and input. That is, the pointer 71 continues to move in the upward direction during the period in which the ending is extended and input, and the window 72 is expanded during the period in which the ending is extended and input. Continue.

As described above, according to the screen control device of the second embodiment, it is possible to perform a predetermined process for displaying the screen by a predetermined amount by voice, as in the first embodiment. Also, the second
In the screen control device of the embodiment, the processing instruction is given by a word and the processing amount is designated by the length of the word ending. Therefore, the processing instruction can be designated by an arbitrary word irrespective of the phoneme to be extended. That is, the number of processes that can be instructed can be set infinitely, and it is possible to handle complicated processes and various types of processes. Further, in the screen control device of the second embodiment, since the process can be designated by uttering a word, the process can be instructed using the word indicating the content of the process, and the instruction in the ordinary language expression can be given. Nearly, a screen control device with an excellent human interface can be realized.

The screen control device of the present invention is the first
The present invention is not limited to the embodiment and the second embodiment, and various modifications can be made. For example, in the above-described embodiment, the method of analyzing the input speech to extract the features and identify the phoneme is to perform a fast Fourier transform (FFT) of 256 sample points based on the input of 16 channels of mel scale, and obtain the result. It was input to a neural network to identify phonemes. However, the present invention is not limited to this, and various widely known analysis methods and signal processing methods may be applied. For example, DTW (Dy (Dy) using multi-template which is often used in speech recognition.
native Time Warping), HMM (Hidden Markov model) which is a statistical method,
Furthermore, various other phoneme analysis techniques may be used.

Further, in the method of detecting the continuation of endings in the second embodiment, the energy level of the input voice signal is measured by the voice input detection unit, and after the time when the word is identified based on the measurement result. The presence or absence of the voice input of was detected, and the continuation of the ending was detected by this. However, a means equivalent to the phoneme continuation detection unit of the first embodiment may be provided in the subsequent stage of the word identification unit to detect the continuation of the ending of a word. FIG. 8 shows states of phoneme detection, word identification, and word ending in that case. As shown in FIG. 8, in such a method, when the word identifying unit recognizes the word "up" in the frame 18, the phonemes identified in the 19th frame and thereafter are compared, and the ending "e" is continued up to the frame 32. It is detected that the phoneme of continues. Then, in the frame 33, when the phoneme is changed while the energy level of the input voice signal is maintained, this method detects the change in the phoneme, and the processing by the word “up” is performed in the frame 32. Can be terminated.

Further, as the display device to which the screen control device of the present invention is applied, a normal display device such as a workstation or a personal computer is illustrated in FIGS. 4 and 7, but the present invention is not limited to this and, for example, It may be a display device such as an in-vehicle television or a car navigation system. In such an environment where manual input is inconvenient, the screen control device of the present invention is more suitable.

Further, in the processing corresponding to the recognized voice signal, in the above-mentioned first and second embodiments, the pointer is moved in the vertical and horizontal directions and the window is enlarged / reduced. . However, various screen operations other than this may be performed. For example, the processing may be such that a specific object on the display screen is rotationally moved, or the entire screen may be scrolled vertically and horizontally, and the entire screen may be enlarged / reduced.

[0043]

According to the screen control device of the present invention, it is possible to directly and intuitively instruct the processing and the processing amount by voice so that the display screen is in a desired state while watching the screen. A screen control device with a good human interface can be provided. Further, according to the screen control method of the present invention, it is possible to provide a screen control method for instructing both the type of processing and the processing amount by voice to the display device.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of a screen control device according to a first embodiment of the present invention.

FIG. 2 is a diagram showing a configuration of a neural network which is a phoneme identification unit of the screen control device shown in FIG.

FIG. 3 is a diagram for explaining detection of phoneme continuation in a phoneme continuation detection unit of the screen control device shown in FIG. 1.

FIG. 4 is a diagram illustrating an operation of a display screen of a display device controlled by the screen control device shown in FIG.

FIG. 5 is a block diagram showing a configuration of a screen control device according to a second embodiment of the present invention.

6 is a diagram illustrating phoneme detection, word identification, and word ending continuation in the screen control device illustrated in FIG. 5.

FIG. 7 is a diagram illustrating an operation of a display screen of a display device controlled by the screen control device shown in FIG.

FIG. 8 is a diagram illustrating detection of an audio signal in a modified example of the screen control device of the second embodiment.

[Explanation of symbols]

10 ... Screen control device 11 ... Acoustic feature extraction unit 12 ... Phoneme identification unit 13 ... Phoneme continuation detection unit 14 ... Screen control unit 50 ... Screen control device 51 ... Acoustic feature extraction unit 52 ... Phoneme identification unit 53 ... Word identification unit 54 ... Screen control unit 55 ... Voice input detector

─────────────────────────────────────────────────── ─── Continuation of front page (51) Int.Cl. ⁷ Identification code FI theme code (reference) G10L 15/18 G10L 3/00 537H 15/28 551P 551Q F term (reference) 5B050 CA07 EA12 FA02 FA10 5D015 BB02 KK01 KK02 5E501 AA01 AA22 AA23 BA20 CA01 CB01 FA06

Claims (2)

[Claims] 1. A screen control device for enlarging or reducing an image displayed on a screen according to the contents of a word of a voice registered in advance, and enlarging or reducing it according to the duration of the ending of the voice. A voice recognition means for recognizing the content of the word of the input voice and determining whether to enlarge or reduce the image displayed on the screen, and a duration detecting means for detecting the duration of the ending of the input voice. And a screen control device for enlarging or reducing the image displayed on the screen according to the detected duration of the ending. 2. A screen control method for enlarging or reducing an image displayed on a screen according to the content of a voice word registered in advance, according to the duration of the ending of the voice. And a voice recognition step of recognizing the content of a word of the input voice to determine enlargement or reduction of the image displayed on the screen, and a duration detection for detecting the duration of the ending of the input voice. A screen control method comprising: a step; and a screen control step of enlarging or reducing an image displayed on the screen according to the detected duration of the ending.