JP3917101B2 - Mobile phone terminal and voice level control program - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a method for controlling an input voice level in a system in which a voice recognition function is not performed in a mobile phone but performed on the center side via a network in an information service providing system using voice input from a mobile phone by a user. .
[0002]
[Prior art]
In general, a speech recognition system has an appropriate input speech level range that maximizes recognition performance. If the level is too high, the voice will be distorted. If the level is too low, it will be difficult to separate the noise and the voice. In either case, it will not be possible to accurately extract the voice feature amount, and the voice recognition rate will deteriorate. Therefore, a device is known in which the peak level of speech is detected at the input stage of the speech recognition device, and the amplification gain is dynamically adjusted so as not to exceed the maximum input level that can be handled by the speech recognition device (Patent Document 1). . In addition, there is also known an apparatus in which a function for encoding and temporarily capturing voice is provided in the front stage of the voice recognition apparatus, and the unit is set to an optimum constant gain for the voice recognition unit during the unit utterance (Patent Document). 2, 3).
[0003]
[Patent Document 1]
JP 05-265484 A [Patent Document 2]
JP 2001-117585 A [Patent Document 3]
Japanese Patent Laid-Open No. 2002-091487
[Problems to be solved by the invention]
However, in the apparatus that performs dynamic gain adjustment by detecting the voice peak level and noise level shown in Patent Document 1, the gain itself is adjusted in real time even during the utterance, so the level itself is almost ideal. However, there is a problem that the linearity of the signal amplitude is not guaranteed and the speech recognition rate is not necessarily improved. In addition, in the devices that perform gain adjustment for each unit utterance by temporarily capturing voices shown in Patent Documents 2 and 3, since gain adjustment is performed for each utterance, linearity of the voice amplitude is ensured within the unit utterance. There is a problem that voice transmission delay occurs due to temporary buffering of the voice signal, and deterioration of the response, which is important when a natural conversation is performed using the voice recognition system, cannot be avoided. Furthermore, as a point common to these apparatuses, it is assumed that the level adjustment is performed immediately before the voice recognition process, and the case where the level adjustment and the voice recognition are connected through the mobile phone network is not assumed. In the codec of a mobile phone, there is a mechanism for improving the SN ratio by treating the noise as noise when the noise is below a certain level. On the other hand, if the noise exceeds that level, it is handled as speech, so that even if it is large, it is transmitted in a state where the S / N ratio of speech is very degraded, and there is a problem that speech recognition rate deteriorates.
[0005]
The present invention has been made in view of such circumstances, and provides a mobile phone terminal and a voice level control program capable of performing appropriate voice level adjustment when performing voice recognition processing using a mobile phone terminal. The purpose is to provide.
[0006]
[Means for Solving the Problems]
The present invention is a mobile phone terminal including an encoding unit that encodes a voice signal from a microphone and transmits the encoded signal to a mobile telephone network, and detects a minimum level value within a predetermined time of the voice signal output from the microphone. A level monitoring unit that outputs the minimum level value at each predetermined time interval, and an appropriate gain is obtained by a predetermined calculation formula based on the minimum level value output by the level monitoring unit, and the sound is obtained using the appropriate gain. And a gain adjusting means for adjusting the gain of the signal and outputting it to the encoding means.
[0007]
The present invention is characterized in that the gain adjusting means obtains the appropriate gain by adding a margin according to an assumed amount of change in noise level.
[0008]
The present invention relates to an audio level control program that operates in a mobile phone terminal provided with an encoding means for encoding an audio signal from a microphone and transmitting it to a mobile telephone network, and within a predetermined time of the audio signal output from the microphone A level monitoring process for detecting the lowest level value at each predetermined time interval and outputting the lowest level value at each predetermined time interval; and, based on the lowest level value output by the level monitoring process, obtaining an appropriate gain by a predetermined calculation formula, The gain of the audio signal is adjusted using an appropriate gain, and the computer performs gain adjustment processing to output to the encoding means.
[0009]
The present invention is characterized in that the gain adjustment processing obtains the appropriate gain by adding a margin according to an assumed amount of change in noise level.
[0010]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, a mobile phone terminal according to an embodiment of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram showing the configuration of the embodiment. In this figure, reference numeral 1 denotes a mobile phone terminal. Reference numeral 2 denotes a microphone that collects sound and converts it into an audio signal, and is a microphone that is provided in a mobile phone terminal for making a call. Reference numeral 3 denotes a level monitoring unit that monitors the level of an audio signal output from the microphone 2. Reference numeral 4 denotes a gain adjusting unit that adjusts the gain of the audio signal. Reference numeral 5 denotes an encoding unit that encodes and outputs an audio signal. Reference numeral 6 denotes a mobile telephone network. Reference numeral 7 denotes a decoding unit that is provided in the network 6 and decodes the signal encoded by the encoding unit 5. Reference numeral 8 denotes a center system that provides a service. The service provided by the center system 8 is, for example, a service for making a ticket reservation for transportation using the mobile phone terminal 1. Reference numeral 9 denotes a CTI (Computer Telephony Integration) unit that connects the center system 8 to the network 6. Reference numeral 10 denotes a speech recognition unit that recognizes speech collected by the microphone 2. Reference numeral 11 denotes a processing unit that performs processing such as ticket reservation based on the output of the speech recognition unit 10.
[0011]
Here, with reference to FIG. 3, the outline | summary of operation | movement of the mobile telephone terminal 1 shown in FIG. 1 is demonstrated. First, the level monitoring unit 3 always monitors the input level of the gain adjusting unit 4. Then, the minimum input level in the time interval is detected every predetermined time (for example, 3 to 5 seconds), and the level information is notified to the gain adjusting unit 4. The gain adjusting unit 4 sets a level obtained by adding a certain margin to the notified level as a noise level in the time interval. The reason for adding the margin is that the ambient noise level generally fluctuates within a certain range, and if the minimum value of noise is extracted, the maximum value of the noise level may be overestimated. Subsequently, when a signal of this noise level is input, the gain adjusting unit 5 causes the signal output from the gain adjusting unit 4 to the encoding unit 5 to be equal to or lower than a reference level determined as sound by the encoding unit 5. adjust. Then, the level monitoring unit 3 sends level information to the level gain adjusting unit 4 and simultaneously resets the value of the minimum input level. Thus, the level monitoring unit 3 detects the noise level every predetermined time (for example, 3 to 5 seconds), and the gain adjusting unit 4 repeats the operation of adjusting the gain each time.
[0012]
Next, details of the operations of the level monitoring unit 3 and the gain adjusting unit 4 will be described with reference to FIG. First, the level monitoring unit 3 detects the lowest level of the audio signal level within a predetermined time (for example, 3 to 5 seconds) (steps S1 and S2). Then, the level monitoring unit 3 notifies the gain adjusting unit 4 of the lowest level value within a predetermined time (step S3). The operation (steps S1 to S3) in the level monitoring unit 3 is repeatedly executed. As a result, the level monitoring unit 3 notifies the gain adjustment unit 4 of the minimum level value in the predetermined time interval every predetermined time (for example, 3 to 5 seconds).
[0013]
On the other hand, the gain adjusting unit 4 receives the notification of the lowest level value from the level monitoring unit 3 and calculates an appropriate gain (step S4). The appropriate gain is calculated by equation (3). Here, a method for calculating the appropriate gain will be described.
The minimum input level within a predetermined time is Ln [dB], the margin is M [dB], the initial gain is G0 [dB], the appropriate maximum noise input level (the level monitoring unit 3 immediately before the SN ratio suddenly deteriorates at the gain G0) Input level) is Lin [dB], the appropriate maximum noise output level (the output level of the gain adjusting unit 4 immediately before the SN ratio suddenly deteriorates at the gain G0) is Lout [dB], and the appropriate gain is Gan [dB]. In the initial state, equation (1) is established.
Lin + G0 = Lout (1)
[0014]
Every predetermined time interval, it is necessary to prevent the output from exceeding Lout at the minimum input level in that interval. Since the minimum input level of the immediately preceding section and the minimum input level of the current section slightly change, a predetermined margin M is set. When the margin M is not so large, since the S / N degradation is less when the output level is smaller than when the output level is too large (exceeding Lout), the margin M is set to a positive value. The margin M may be determined in advance according to the assumed amount of change in the noise level. From the above, equation (2) is established in each section.
Ln + Gan + M = Lout (2)
Therefore, the appropriate gain Gan is calculated by the equation (3).
Gan = Lout-Ln-M = G0- (Ln-Lin) -M (3)
[0015]
Next, the gain adjusting unit 4 changes the gain to the appropriate gain Gan obtained by calculation (step S5). As a result, the gain of the audio signal output from the microphone 2 is adjusted to the appropriate gain Gan and output to the encoding unit 5.
[0016]
As described above, the audio signal level input to the encoding unit 5 can always be equal to or lower than the reference level in a state where there is no utterance. Therefore, as long as there is no sudden change in the ambient noise level, it can be expected that the noise level is always kept lower than the reference level of the encoding unit 5 during speech, and the S / N ratio is added to the voice input to the voice recognition process through the network. Therefore, the performance of voice recognition processing can always be maximized. Further, by setting the minimum level detection interval of the level monitoring unit 3 to be equal to or longer than the length of one utterance in the voice conversation service, the amplification factor does not frequently change during one utterance, and output to the encoding unit 5 Therefore, the voice recognition processing performance can be maximized. In addition, since the level can be adjusted without buffering the audio signal from the microphone 2, a delay in the signal does not occur, and a voice response with good response can be realized. Further, the level monitoring unit 3 only needs to perform the process of updating the minimum value, and the gain adjustment is performed by taking out the gain from the level monitoring unit 3 once every few seconds (which changes according to the length of the utterance vocabulary). It is only necessary to perform the processing to be performed. Therefore, it is not necessary to use parts with high processing capability.
[0017]
1 is recorded on a computer-readable recording medium, and the program recorded on the recording medium is read into a computer system and executed to adjust the gain. Processing may be performed. Here, the “computer system” includes an OS and hardware such as peripheral devices. The “computer-readable recording medium” refers to a storage device such as a flexible medium, a magneto-optical disk, a portable medium such as a ROM and a CD-ROM, and a hard disk incorporated in a computer system. Further, the “computer-readable recording medium” refers to a volatile memory (RAM) in a computer system that becomes a server or a client when a program is transmitted via a network such as the Internet or a communication line such as a telephone line. In addition, those holding programs for a certain period of time are also included.
[0018]
The program may be transmitted from a computer system storing the program in a storage device or the like to another computer system via a transmission medium or by a transmission wave in the transmission medium. Here, the “transmission medium” for transmitting the program refers to a medium having a function of transmitting information, such as a network (communication network) such as the Internet or a communication line (communication line) such as a telephone line. The program may be for realizing a part of the functions described above. Furthermore, what can implement | achieve the function mentioned above in combination with the program already recorded on the computer system, and what is called a difference file (difference program) may be sufficient.
[0019]
【The invention's effect】
As described above, according to the present invention, it is possible to obtain an effect that the audio signal level input to the encoding unit can always be equal to or lower than the reference level when there is no utterance. Therefore, as long as there is no sudden change in the ambient noise level, it can be expected that the noise level is always kept lower than the reference level of the encoding unit even during speech, and the S / N ratio of the voice input to the voice recognition process through the network is No degradation occurs, and the performance of the speech recognition process can always be maximized.
[Brief description of the drawings]
FIG. 1 is a block diagram showing a configuration of an embodiment of the present invention.
FIG. 2 is an explanatory diagram showing an outline of a level adjustment operation.
FIG. 3 is a flowchart showing an operation of the mobile phone terminal 1 shown in FIG. 1;
[Explanation of symbols]
DESCRIPTION OF SYMBOLS 1 ... Mobile phone terminal 2 ... Microphone 3 ... Level monitoring part 4 ... Gain adjustment part 5 ... Encoding part 6 ... Network 7 ... Decoding part 8 ... Center system 9 ... CTI part 10 ... Voice recognition part 11 ... Processing part

Claims (4)

A cellular phone for transmitting a voice signal from a microphone to a mobile telephone network via an encoding means for encoding , and causing a voice recognition device connected to the mobile telephone network to perform voice recognition of the voice collected by the microphone. A terminal,
A level monitoring means for detecting a minimum level value L n at a predetermined time of the audio signal, and outputs the outermost low-level value L n for each said predetermined time interval which is output from the microphone,
Based on the minimum level value L n output from said level monitoring means, a total of formula
G an = G 0 − (L n −L in )
(Where G 0 is a predetermined initial gain, and L in is a predetermined appropriate maximum noise level)
Seek a proper gain G an, makes the appropriate gain using a G an, to adjust the gain of the audio signal for each said predetermined time interval, characterized by comprising a gain adjusting means for outputting to said encoding means mobile Phone terminal. 2. The gain adjusting means obtains the appropriate gain G an by subtracting a margin M, which is a positive value corresponding to an assumed amount of change in noise level, from the appropriate gain G an. The mobile phone terminal described in 1. A cellular phone for transmitting a voice signal from a microphone to a mobile telephone network via an encoding means for encoding , and causing a voice recognition device connected to the mobile telephone network to perform voice recognition of the voice collected by the microphone. An audio level control program that runs on a terminal,
A level monitoring process detects the minimum level value L n, and outputs the outermost low-level value L n for each of the predetermined time intervals in a predetermined time period of the audio signal output from the microphone,
Based on the minimum level value L n output by the level monitoring process, total formula
G an = G 0 − (L n −L in )
(Where G 0 is a predetermined initial gain, and L in is a predetermined appropriate maximum noise level)
Seek a proper gain G an, by the feature that the proper gain with reference to G an, to adjust the gain of the audio signal for each said predetermined time interval, to perform a gain adjustment processing to be output to the encoding unit to the computer Voice level control program. 4. The gain adjustment process is characterized in that the appropriate gain G an is obtained by subtracting a margin M which is a positive value corresponding to an assumed amount of change in noise level from the appropriate gain G an. The sound level control program described in 1.

Images