JP2000010577A - Voiced sound/voiceless sound judging device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a voiced sound/voiceless sound judging device capable of rapidly and correctly judging a voiced sound or a voiceless sound and being constituted with a simple circuit. SOLUTION: This voiced sound/voiceless sound judging device is constituted of a microphone 11 for inputting a voice, an amplifier 12 for amplifying the signal from the microphone 11 to prescribed amplitude, an A/D converter 13 for converting the signal from the amplifier 12 to a digital signal, a low-pass filter 14 for extracting the signal of a low band from the digitized signal, an amplitude average measuring instrument 15 for measuring the average of the amplitude in the prescribed time interval of the extracted low band signal, a number of zero-crossing counter 16 for measuring the number of zero-crossing of the digitized signal and a voiced sound/voiceless sound discriminator 17 for discriminating the voiced sound from the voiceless sound based on the outputs of the amplitude mean measuring instrument 15 and the number of zero-cross counter 16 and deciding the ratio of the voiceless sound to the input voice.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a voiced / unvoiced sound judging apparatus for accurately and easily judging the ratio between voiced sound and unvoiced sound contained in a voice input.

[0002]

2. Description of the Related Art The following is known as a voiced / unvoiced sound discriminating apparatus.

[0003] The first is that a voice input from a microphone is
The signal is converted into a digital signal via an A / D converter, and the number of times the amplitude level of the signal crosses the zero level at a fixed time interval, that is, the number of zero crossings is measured. If the number of zero crossings falls below the first threshold with respect to the first threshold and the second threshold (the first threshold <the second threshold) set inside the apparatus, it is determined that the voiced sound is present. The voiced / unvoiced sound determination device determines that the voiced sound is unvoiced when the value exceeds the second threshold value, and that the voiced sound and the unvoiced sound are mixed when the value is between the first threshold value and the second threshold value.

However, this apparatus has an advantage that it can be constituted by a relatively simple circuit. However, in the case of a voice in which a noise component is added to a periodic voiced sound, for example, a voice such as a voiced consonant, a zero amplitude level is obtained. The value of the number of zero crossings is increased due to the influence of noise components in the vicinity, and this voice may be erroneously determined to be unvoiced.

[0005] The second is to convert the voice input from the microphone into a digital signal via an A / D converter,
The autocorrelation function is calculated at regular time intervals. This autocorrelation function indicates the degree of periodicity of the voice,
If there is a clear peak in the autocorrelation function value, it is determined that the voiced sound has a strong periodicity. On the other hand, if there is no clear peak, it is determined that the voiced sound has a poor periodicity.

However, in this apparatus, voiced sound / unvoiced sound can be accurately determined because the periodicity of speech is used as a basis for determination. However, a large number of multiply-accumulate operations are required for calculating an autocorrelation function. Therefore, the circuit configuration becomes complicated,
Also, the calculation requires time.

[0007]

SUMMARY OF THE INVENTION Accordingly, an object of the present invention is to provide a voiced / unvoiced sound discriminating apparatus which can quickly and accurately determine a voiced sound and an unvoiced sound, and which can be constituted by a simple circuit.

[0008]

SUMMARY OF THE INVENTION The present invention has been made in view of the above problems, and has a counting means for counting the number of zero crossings of an input audio signal, and a low-pass filter for converting the input audio signal. After passing, an average value of the amplitude of the passed audio signal is measured at predetermined time intervals, and a determination unit that determines voiced sound and unvoiced sound based on the average value of the amplitude and the number of zero crossings is provided. A voiced / unvoiced sound determination device is configured.

Further, the voiced / unvoiced sound judging device has a structure in which a judgment standard for voiced sound and unvoiced sound set in the judging means is automatically set to a predetermined value by an average value. Resolve.

A criterion for determining the ratio of unvoiced sounds is determined according to the amplitude of the input voice signal, and the ratio of unvoiced sounds in the input voice is determined based on the number of zero crossings of the input voice and the criterion.

[0011]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention relates to a voiced / unvoiced sound determination device capable of accurately and easily determining the degree of voiced sound and unvoiced sound contained in an input voice signal. A criterion for determining the ratio of unvoiced sounds included in the input voice is changed according to the magnitude of the amplitude, and the ratio of unvoiced sounds in the input voice is determined according to the criterion.

Next, an embodiment will be described with reference to FIGS.
This will be described with reference to FIG. Here, FIG. 1 is a block diagram of a voiced / unvoiced sound discriminating apparatus according to the present invention, and FIG. 2 is a diagram showing an input to a voiced / unvoiced discriminator and the number of zero crossings, and an output of the voiced / input voice ratio UR. (Unvoiced Sound Rati
It is a figure which shows the relationship with o). FIG. 3 is a diagram showing the relationship between the number of zero crossings and the unvoiced sound / input voice ratio UR when the threshold value in the voiced / unvoiced sound discriminator is changed based on the average amplitude value from the amplitude average measuring device.

First, as shown in FIG. 1, a voiced / unvoiced sound judging device 1 includes a microphone 11 for inputting a voice, a microphone 11
12, an A / D converter 13 for converting a signal from the amplifier 12 into a digital signal, and a low-pass filter for extracting a low-frequency signal from the digitized signal. 14; an amplitude average measuring device 15 for measuring the average of the amplitude of the extracted low-frequency signal in a predetermined time interval; and a zero crossing for measuring the number of crossings of the digitized signal with the zero level, that is, the number of zero crossings. A voiced / unvoiced sound discriminator 17 that discriminates between voiced sounds and unvoiced sounds based on the outputs of the number counter 16 and the amplitude average measuring device 15 and the zero-crossing number counter 16, and determines the ratio of unvoiced sound to input sound. It is configured.

Next, the operation of the voiced / unvoiced sound discriminating apparatus 1 having the above configuration will be described. The voice including the voiced sound and the unvoiced sound input from the microphone 11 is amplified by the amplifier 12 and then converted to a digital signal via the A / D converter 13. The signal output from the A / D converter 13 is split into two, one of which is sent to a low-pass filter 14,
The other is sent to the zero-crossing counter 16.

The signal sent to the low-pass filter 14 is
The high-frequency component that is the unvoiced sound component included in the signal is removed, and as a result, only the low-frequency component that is the voiced sound component in the voice is output. Next, the signal from the low-pass filter 14 is applied to an amplitude average
And an average amplitude value per predetermined time is measured. Due to the nature of the input voice, the average amplitude value is a large value when the input voice is close to a voiced sound, and is a small value when it is unvoiced or close to no voice.

In the signal sent from the A / D converter 13 to the zero-crossing number counter 16, the number of zero-crossings per predetermined time included in the signal is counted. The zero crossing means that a signal crosses a zero level from positive to negative and from negative to positive, and the number of these crossings within a predetermined time is counted and defined as the number of zero crossings. Therefore, when there are many unvoiced sound components in the input voice, that is, when there are many high-frequency noise components, the number of times the signal crosses the zero level increases,
Their count, the number of zero crossings, takes a large value, while
When there are many voiced sound components in the input voice, that is, when there are many low-frequency components, the number of times the signal crosses the zero level decreases, and the number of zero crossings becomes a small value.

The calculated number of zero crossings is input to a voiced / unvoiced sound discriminator 17. Thresholds n1 and n2 (n1 <n2) of the number of zero-crossings are set in the voiced / unvoiced sound discriminator 17, and the thresholds n1 and n2 are compared with the input number of zero-crossings to determine whether the voiced sound or unvoiced sound is present. An unvoiced sound is determined. As a result of the determination, the ratio of unvoiced sound (U
R) is output. The ratio (UR) is 0.0 to 1.0.
Take the value of Further, the threshold values n1 and n2 are appropriately changed according to the average amplitude value of the input audio signal, and form a feature of the present invention as will be described later in detail.

Next, the discrimination between a voiced sound and an unvoiced sound will be described. The horizontal axis shown in FIG. 2 is the number of zero crossings, and the vertical axis is the ratio (U
The graph R) is used for discriminating voiced sounds and unvoiced sounds. Actually, this graph is stored in the voiced / unvoiced sound discriminator 17 as a table.

In the graph shown in FIG. 2, the input voice is a voiced sound region from a completely voiced sound to the number of zero crossings = n1, an intermediate region from the number of zero crossings = n1 to n2, a complete unvoiced sound from the number of zero crossings = n2. Up to the unvoiced sound area.

When the number of zero crossings of the input voice is in the voiced sound area, the ratio (UR) = 0.0 is output from the voiced / unvoiced sound discriminator 17, and it is determined that the input voice is only the voiced sound. You. When the number of zero crossings of the input voice is in the unvoiced sound area, the ratio (UR) = 1.0 is output, and it is determined that the input voice is only the unvoiced sound. When the number of zero crossings of the input voice is in the intermediate region, the ratio (UR) takes a value between 0.0 and 1.0, and it is determined that the input voice includes voiced and unvoiced sounds. . At this time, the ratio (UR) is given by the ratio between the input voice and the unvoiced sound included in the input voice. For example, in FIG. 2, when the number of zero crossings = ni, the ratio (UR) = 0.
Assuming that 45 is output from the voiced / unvoiced sound discriminator 17, 45% of the input voice is unvoiced and the remaining 55
% Is determined to be voiced.

The voiced / unvoiced sound determination device 1 of the present invention
Is provided with an amplitude average measuring device 15 for measuring an average of amplitudes of input voices at predetermined time intervals. The average amplitude value of the input voice signal output from the amplitude average measuring device 15 is input to the voiced / unvoiced sound discriminator 17, and the above-mentioned threshold values n1 and n2 are appropriately changed according to the magnitude of the average amplitude value. is there. This is because when the average amplitude value increases, voiced sounds increase in the input voice, and when the average amplitude value decreases, unvoiced sounds increase.

FIG. 3 shows that the average amplitude value increases and the threshold value n
1 is a graph when n2 is changed to n1a and n2a in a direction in which the number of zero crossings increases. In this case, the input voice is a voiced sound area from a completely voiced sound to the number of zero crossings = n1a, an intermediate area from the number of zero crossings = n1a to n2a, and an unvoiced sound area from the number of zero crossings = n2a to a completely unvoiced sound. Therefore, if the input voice is a completely voiced sound and the number of zero crossings = n1a, the ratio (U
R) = 0.0 is output and it is determined that the input voice is only a voiced sound, and the ratio (UR) = 1.0 is output from the input voice from the number of zero crossings = n2a to the complete unvoiced sound. It is determined that the input voice is only unvoiced sound.

Now, looking at the case where the number of zero crossings = ni, FIG. 2 shows that n1 <ni <n2, and the input speech at this time is determined to have a voiced sound and an unvoiced sound at a predetermined ratio. On the other hand, in FIG. 3, ni <n1a <n2a
In this case, even in the case of the same ni, the input voice is determined to be only a voiced sound. Conversely, if the threshold value is n1a <n2a <ni, the input voice is determined to be only unvoiced sound.

As described above, the threshold value n depends on the average amplitude value.
By changing 1 and n2, voices that can be judged to be unvoiced only by the number of zero crossings, such as voiced consonants with high-frequency noise components added to low-frequency periodic components, are accurately obtained. It is possible to judge the voice sound.

The ratio (U) determined as described above
R) is used for various processes of the input audio signal. For example, when voiced sound is treated as meaningful data in the input voice, unvoiced sound is unnecessary, and the unvoiced sound portion is excluded from processing based on the obtained ratio (UR), thereby shortening processing time and increasing efficiency. It also reduces the load on the device and contributes to the weight reduction of the device.

The table for determining the UR based on the number of zero-crossings is not limited to the above-mentioned two tables, but may be a larger table. Further, the threshold of the table may be automatically and continuously changed according to the amplitude of the input voice, and the ratio (UR) may be determined based on the threshold based on the relationship with the number of zero crossings.

The present invention is not limited to the configuration of the above-described embodiment, but may be any configuration that embodies the technical idea of the present invention.

[0028]

As described above in detail, according to the voiced / unvoiced sound judging device of the present invention, a voice having a noise component added to a voiced sound having a periodicity, for example, a voice such as a voiced consonant. The voiced / unvoiced sound can be accurately determined in a short time with a simple circuit without erroneously determining the unvoiced sound.

[Brief description of the drawings]

FIG. 1 is a block diagram of a voiced / unvoiced sound determination device according to an embodiment of the present invention.

FIG. 2 is a diagram illustrating a relationship between the number of zero crossings as an input to a voiced / unvoiced sound discriminator and an unvoiced sound / input voice ratio UR as an output.

FIG. 3 is a diagram showing the relationship between the number of zero-crossings and the unvoiced sound / input voice ratio UR when the threshold in the voiced / unvoiced sound discriminator is changed based on the average amplitude value from the amplitude average measuring device.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 ... Voiced / unvoiced sound determination apparatus, 11 ... Microphone, 12 ... Amplifier, 13 ... A / D converter, 14 ... Low-pass filter,
15: Amplitude average measuring instrument, 16: Zero-crossing counter, 17 ...
Voiced / unvoiced sound discriminator

Claims (2)

[Claims] 1. A counting means for counting the number of zero crossings of an input audio signal, and after passing the input audio signal through a low-pass filter, an average value of the amplitude of the passed audio signal is determined by a predetermined value. A voiced / unvoiced sound determination device, comprising: a measuring means for measuring at time intervals; and a determining means for determining a voiced sound or an unvoiced sound based on the average value of the amplitude and the number of zero crossings. 2. The apparatus according to claim 1, wherein a criterion for determining whether a voiced sound or an unvoiced sound is set in said determining means is automatically set to a predetermined value based on said average value. A voiced / unvoiced sound determination device as described in the above.