JPH07336289A - Vox control communication device - Google Patents

Abstract

PURPOSE:To eliminate the cutting of a talk at its beginning and obtain high quality as to the communication device which employs VOX control. CONSTITUTION:A speech signal is converted into a digital signal by a PCM encoder 1 and the mean power and reflection coefficient of its current frame are calculated by a speech encoder 2 and inputted to a speech detector 3. The speech detector 3 calculates the quantity of variation in mean power between a voiceless section and the current frame. Further, a final threshold value is determined from the weighted mean of the threshold value calculated from the mean power of the voiceless section and the threshold value calculated from the mean predicted gain of the voiceless section and the predicted gain of the current frame and this threshold value is compared with the variation quantity to decides whether or not the current frame is a voiced or voiceless section. The mean power and mean predicted gain of the voiceless section are updated in order with the weighted mean of the mean power and predicted gain of the current frame decided as the voiceless section.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a VOX control communication device, and more particularly to a VOX control used in a digital mobile communication device.

[0002]

2. Description of the Related Art Conventionally, VOX (Voice Operated Transistor) has been used as a method for saving power consumption during transmission of radio equipment.
mitter) is known. The VOX is a technique of detecting an input voice signal and turning on the power of the transmitter only when the voice signal is present, and turning it off when the voice signal is not present.

In particular, in mobile communication devices such as mobile phones, since the power is supplied by a battery, this VOX is widely used as a power saving technique.

In the VOX, if the rising time from the detection of the voice signal to the turning on of the electric power of the transmitter is too long, so-called head disconnection occurs and the voice becomes difficult to hear on the receiving side. Therefore, it is extremely important to detect the presence / absence of a voice signal.

Therefore, in the conventional VOX, whether the voice is present or not is determined by setting a fixed threshold level for the input voice level and comparing these values.

[0006]

However, in the configuration for determining whether there is voice or no voice by using the fixed threshold value as described above, when a low-level voice signal such as a consonant is input, However, it may be erroneously determined to be smaller than the fixed threshold value, and as a result, there is a problem that the speech is missing and the sound quality deteriorates.

Therefore, the applicant of the present invention previously filed Japanese Patent Application No. 6-41.
In No. 1, we proposed a configuration that determines the threshold value based on the average power in the silent section, and compares it with the variable threshold value to determine whether there is sound or silence. In this way, by determining the threshold value for determining whether there is sound or no sound according to the average power of the silent section, it is possible to reliably detect a low level audio signal even under high noise.

On the other hand, even when the threshold value is determined in accordance with the average power in the silent section as described above, an erroneous determination may occur when a specific consonant (for example, a sa-syaku) comes to the beginning of the talk. , There was a problem that it was not sufficient to prevent the head disconnection for all audio signals.

The present invention has been made in view of the above problems of the prior art, and an object of the present invention is to reliably prevent the beginning of a talk under high noise and to perform highly efficient VOX control with less malfunction. It is to provide a VOX control communication device.

[0010]

In order to achieve the above object, a VOX control communication apparatus according to claim 1 determines whether a current frame is voiced or silenced when encoding an analog input voice. In a VOX control communication device that transmits only during sound, a change amount calculation unit that calculates a change amount of the average power of the current frame with respect to the average power of a silent section, and a predicted gain based on at least first-order and second-order reflection coefficients Predictive gain calculating means, average calculating means for calculating an average predictive gain in a silent section, first threshold calculating means for calculating a first threshold value based on the average power of the silent section, and the average predictive gain. And second based on the prediction gain of the current frame
Second threshold value calculating means for calculating a threshold value, and the first
And a third threshold value calculating means for calculating a third threshold value by a weighted average of the second threshold value, and whether the current frame is voiced or silenced by comparing the change amount and the third threshold value. When the current frame is determined to be silent by the determining means, the average power of the silent section is updated by a weighted average of the average power of the silent section and the average power of the current frame determined to be silent. In addition, there is provided an updating means for updating the average prediction gain of the silent section with a weighted average of the average prediction gain of the silent section and the prediction gain of the current frame determined to be silent.

In order to achieve the above object, a VOX control communication device according to a second aspect is a VOX control device according to the first aspect.
In the control communication device, the predictive gain calculating means calculates the predictive gain based on the reflection coefficients from the first order to the fourth order.

Further, in order to achieve the above object, the VOX control communication device according to claim 3 is the VOX control communication device according to claim 1 or 2, wherein the first threshold value calculating means is The first threshold value is calculated so as to be inversely proportional to the average power in the silent section.

Further, in order to achieve the above object, the VOX control communication device according to claim 4 is the VOX control communication device according to claim 1, claim 2 or claim 3, in which the second threshold value is calculated. When the average prediction gain is Rav, the prediction gain is R, and positive constants are A and B,

The Equation 1] ^{TH = (A / Rav 3)} · R 3 -B and calculates the second threshold value TH.

[0014]

In the VOX control communication device of the present invention, the lack of consonants in the case of determining whether there is voice or no voice only with the threshold value (first threshold value) calculated based on the average power of the silent section is prevented. Therefore, the final threshold value is determined by weighted averaging with the threshold value (second threshold value) based on the prediction gain calculated based on the reflection coefficient, and the final threshold value is compared with the change amount. Determine whether it is sound or silence. At this time,
The function used in determining the second threshold value is adjusted based on the average prediction gain of the silent period, and the second threshold value is determined based on the average prediction gain of the silent period and the prediction gain of the current frame. become.

When a consonant is input, attention is paid to the fact that the prediction gain calculated based on the reflection coefficient decreases, and the consonant is determined by determining the threshold value (second threshold value) based on the prediction gain. It can be reliably detected. Here, if the threshold value (second threshold value) is simply determined so as to have a positive correlation with the prediction gain, the second threshold value may unnecessarily increase as the prediction gain increases. Therefore, the second
The second threshold value is set to an appropriate value by using a coefficient having a negative correlation with the average prediction gain in the silent section as the coefficient of the function for determining the threshold value. On the other hand, by determining the threshold value (first threshold value) based on the average power in the silent section, it is possible to cope with a minute power fluctuation even under high noise. Therefore, by determining the final threshold value (third threshold value) by the weighted average of these first threshold value and second threshold value, it is possible to reliably detect consonants even under high noise. Is possible.

The first threshold value is set so as to be inversely proportional to the average power of the silent section, that is, the threshold value is lowered when the noise is high and the average power of the silent section is large. The predictive gain in calculating the threshold is at least the first and second order reflection coefficients, preferably from the first to the fourth.
The reflection coefficients up to the next are used. This is because the threshold value is set to an optimum value in a short time without excessively responding to fluctuations in the reflection coefficient.

[0017]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT An embodiment of a VOX control communication device of the present invention will be described below with reference to the drawings. FIG. 1 shows a block diagram of the configuration of this embodiment. A voice signal input from a microphone (not shown) is input to the PCM encoder 1 and converted into a digital signal. The digital signal from the PCM encoder 1 is input to the speech encoder 2. The voice encoder 2 calculates the average power and reflection coefficient of the current frame. The calculated average power and reflection coefficient are input to the voice detector 3. The voice detector 3 performs a predetermined calculation and a size comparison described later to determine whether the current frame is voiced or not, and outputs a control signal to the wireless unit 4.
The wireless unit 4 supplies a digital signal to the antenna 5 for transmission when the current frame has a sound based on the control signal from the voice detector 3.

FIG. 2 shows a flowchart of the processing performed by the voice detector 3. As described above, the speech detector 3 receives the average power of the current frame from the speech encoder 2, but the speech detector 3 receives the average power of the inputted current frame and the calculated silent period. The amount of change from the average power is calculated (S101). The calculation of the average power in the speech encoder 2 is performed by calculating the sum of squares of the respective digital signals input from the PCM encoder 1. Further, the average power of the silent section is calculated and sequentially updated based on the average power of the section already determined to be silent, the details of which will be described later. After the change amount of the average power of the current frame with respect to the average power of the silent section is calculated, the voice detector 3
Calculates the prediction gain based on the first to fourth order reflection coefficients input from the speech encoder 2 (S102). Prediction gain R is when the reflection coefficient is ri (i = 1, 2, 3, 4)

[Equation 2] Is calculated by When the prediction gain is calculated using only the first-order reflection coefficient, the prediction gain fluctuates excessively with respect to the fluctuation of the reflection coefficient, which makes it difficult to optimally determine the threshold value. In addition, when the prediction gain is calculated in consideration of the higher-order reflection coefficients of 5th order or higher, the efficiency decreases because the high-order reflection coefficient hardly contributes and it takes a long time to calculate the prediction gain. There is a problem that ends up. Therefore, when calculating the prediction gain,
It is preferable to use the first to fourth order reflection coefficients for the calculation.

After the predicted gain is calculated based on the first to fourth order reflection coefficients, the voice detector 3 determines the threshold value 1 (first threshold value) based on the average power in the silent section (S1).
03). This threshold value 1 is determined so as to be inversely proportional to the average power in the silent section, and in this embodiment, in particular,

It has been determined by the [number 3] _{TH1 = -0.0833 (P 0 +6)} . Here, TH1 is the threshold value 1, P ₀
Is the average power in the silent section. Note that FIG. 3 shows the relationship between the average power P _{0 in the} silent section and the threshold value 1.

Threshold value 1 based on the average power in the silent section
After the determination is made, the voice detector 3 determines a function for calculating the threshold 2 based on the average value of the prediction gains of the past silent sections calculated in S102 (S104).
In the present embodiment, this threshold value 2 is determined by a cubic function having a positive correlation such that it decreases as the prediction gain decreases and increases as the prediction gain increases, and the third-order coefficient of the function is determined. Is determined by the average prediction gain in the silent section. Specifically, the third-order coefficient Coe is the average prediction gain Rav of the silent section,

## EQU4 ## It is determined by Coe = 6.5 / Rav ³ . In this way, the threshold 2 (second
After the calculation function of (threshold value) is determined, the voice detector 3
Uses the cubic function having the prediction gain R of the current frame and the coefficient Coe determined in S104 to set the threshold value 2 to

[Number 5] is determined by TH2 = Coe · R ³ -1.5. FIG. 4 shows the relationship between the prediction gain and the threshold value 2. Since the third-order coefficient Coe becomes smaller as the average prediction gain in the silent section becomes larger, the threshold value 2 becomes smaller in the end, and conversely, the third-order coefficient Coe becomes larger as the average prediction gain in the silent section becomes smaller. Eventually, the threshold value 2 becomes large.

After the threshold 1 and the threshold 2 are calculated, the final threshold is calculated by the weighted average of these thresholds (S106). In this calculation, each weight is set to 2: 1

[Equation 6] TH = (2TH1 + TH2) / 3

After the final threshold TH is determined, S
The change amount calculated in 101 and the threshold value TH are compared in size (S107). When the amount of change is equal to or greater than the threshold value, it is determined that the current frame has a sound (S108), and a sound signal is output to the wireless unit 4. On the other hand, when the change amount is smaller than the threshold value, a weighted average of the already calculated average power of the silent section and the average power of the current frame is calculated, and the average power of the silent section is updated by this weighted average. (S109),
The average prediction gain of the silent section is updated by the weighted average of the average prediction gain of the silent section and the prediction gain of the current frame (S110). Then, it is determined that the current frame is silent, and a silent signal is output to the wireless unit 4 (S111).

As described above, in the present embodiment, the voice detector 3 determines whether there is sound or no sound based on the amount of change in the average power of the silent section and the average power of the current frame, and determines the average power of the silent section. Since it is updated sequentially, it is possible to cope with a slight power fluctuation under noise and prevent malfunction. Further, as a threshold value used when determining whether there is sound or no sound, a threshold value (threshold value 1) determined based on the average power of the silent section, the average prediction gain of the silent section, and the current frame. Since it is calculated by weighted averaging with a threshold value (threshold value 2) determined based on the prediction gain of, consonant detection is also excellent under colored noise,
It becomes possible to effectively prevent the lack of the talk head and perform high quality communication. That is, the threshold value 1 is determined to have a negative correlation with the average power of the silent section to facilitate detection under high noise, and the threshold value 2 has a positive correlation with the prediction gain. To facilitate detection of consonants, and to determine the coefficient of the decision function of the threshold 2 so as to have a negative correlation with the average prediction gain in the silent section. It is possible to suppress unnecessary increase and optimize the threshold value.

[0024]

As described above, according to the VOX control communication device of the first to fourth aspects, the threshold value for determining whether there is sound or no sound can be optimized even in the presence of noise. It is possible to perform highly efficient VOX control with high quality and less malfunction.

[Brief description of drawings]

FIG. 1 is a configuration block diagram of an embodiment of the present invention.

FIG. 2 is a processing flowchart in the embodiment of the present invention.

FIG. 3 is a graph showing a relationship between average power in a silent section and a first threshold value in the example of the present invention.

FIG. 4 is a graph showing a relationship between a prediction gain and a second threshold value according to the embodiment of the present invention.

[Description of Codes] 1 PCM encoder 2 Speech encoder 3 Speech detector 4 Radio unit 5 Antenna

Claims (4)

[Claims] 1. A VOX control communication device for determining whether a current frame has voice or silence when encoding an analog input voice and transmitting only when there is voice, the average power of the current frame with respect to the average power of a silence period. Change amount calculating means for calculating the change amount of the, a predictive gain calculating means for calculating a predictive gain based on at least first and second order reflection coefficients, an average calculating means for calculating an average predictive gain in a silent section, First threshold value calculating means for calculating a first threshold value based on the average power of the section; and second threshold value calculating means for calculating a second threshold value based on the average prediction gain and the prediction gain of the current frame. A third threshold value calculating means for calculating a third threshold value by weighted averaging the first and second threshold values; and a current frame by comparing the change amount and the third threshold value. A determination unit for determining whether there is sound or silence, and when the determination unit determines that the current frame is silent, the silent section is a weighted average of the average power of the silent section and the average power of the current frame determined to be silent. Update means for updating the average prediction gain of the silent section and a weighted average of the average prediction gain of the silent section and the prediction gain of the current frame determined to be silent, and updating means. VOX control communication device. 2. The VOX control communication device according to claim 1, wherein the predictive gain calculating means calculates the predictive gain based on the reflection coefficients from the first order to the fourth order. 3. The VOX control communication device according to claim 1, wherein the first threshold value calculation means calculates the first threshold value so as to be inversely proportional to the average power of the silent section. A VOX control communication device characterized by the above. 4. Claim 1 or claim 2 or claim 3.
In the VOX control communication device described above, when the average prediction gain of the silent section is Rav, the prediction gain is R, and positive constants are A and B, TH = (A / Rav ³⁾ · R ³ V, characterized in that to calculate the second threshold value TH by -B
OX control communication device.