JP2009065424A - Impulse identification device and impulse identification method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To obtain a device and a method which can efficiently identify voice and impulse sound, in noise determination processing. <P>SOLUTION: An impulse identification device comprising a digital signal processor determining whether an input signal is an impulse, by performing signal operation processing to the signal inputted from an acoustoelectric converter which converts a surrounding sound into an electric signal has a Fourier transformation means that transforms an input signal into frequency characteristics; a first determination means for determining whether a higher harmonic wave is contained in the frequency characteristics; and a second determination means for determining whether the gain level at a predetermined frequency of the frequency characteristics is equal to or higher than a predetermined value. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to an apparatus and a method for discriminating voice and impulse sound in noise determination processing of a noise canceling microphone.

  There is known a noise canceling microphone (hereinafter referred to as “NC microphone”) that discriminates a speaker's voice and ambient noise and reduces (removes) ambient noise included in an input signal and outputs it. Ambient noise mainly includes stationary noise (such as indoor air conditioner sounds and outdoor wind sounds) and sudden sounds (applause, door opening and closing sounds, and hitting objects). Conventionally, sudden sound is identified by a method of identifying a sound having a large sound pressure fluctuation in the time domain of the input signal as a sudden sound (hereinafter referred to as “impulse sound”), or a sound pressure in the low frequency domain of the input signal frequency domain. A method is known in which the ratio of the energy and the sound pressure energy region in the high frequency region is identified as an impulse sound with a predetermined threshold value or more (see, for example, Patent Documents 1 to 3).

Japanese Patent Laid-Open No. 08-288999 Japanese Patent Laid-Open No. 08-87293 JP 01-24665 A

  However, in the conventional identification method as described in Patent Documents 1 to 3, for example, an impulse sound with a small sound pressure level generated when a pen is dropped on a desk has a small sound pressure fluctuation in the time domain. Therefore, it cannot be identified as an impulse sound. In addition, since the sound of the “sa” line (sa, shi, sushi, se, so) has a high ratio of high frequency region components, the ratio of the sound pressure energy in the high frequency region to the low frequency region in the frequency region is large. It will be identified as sound.

  The present invention has been made in view of such a current situation, and can accurately identify even an impulse sound with a low sound pressure level, and if an input sound with a large high frequency range is a voice, the impulse It is an object of the present invention to provide a more accurate impulse identification method that can be prevented from being identified as sound.

  The present invention includes a digital signal processor that performs signal arithmetic processing on an input signal input from an acoustoelectric converter that converts ambient sound into an electric signal, and determines whether the signal is an impulse. An impulse identification device comprising: Fourier transform means for converting the input signal into frequency characteristics; first determination means for determining whether or not harmonics are included in the frequency characteristics; And a second determination means for determining whether or not the frequency gain level is equal to or higher than a predetermined value.

  The present invention also provides Fourier transform means for converting an input signal input from an acoustoelectric converter that converts ambient noise into an electrical signal into frequency characteristics, and whether the frequency characteristics include harmonics. An impulse identification method in a noise canceling device comprising: first determination means for determining; and second determination means for determining whether a gain level of a predetermined frequency of the frequency characteristic is a predetermined value or more, The step of converting the input signal into a frequency characteristic by the Fourier transform unit, the step of determining whether or not a harmonic is included in the frequency characteristic by the first determination unit, and the second determination unit Whether the gain level of the predetermined frequency of the frequency characteristic determined to contain no harmonics in the determination by the first determination means is greater than or equal to a predetermined value Determining a, and characterized mainly in that it has a.

  According to the present invention, it is possible to identify whether the sound is an impulsive sound by identifying the characteristics of the frequency component of the signal input from the microphone and the characteristics of the gain level of the predetermined frequency in the frequency component of the input signal. Therefore, it is possible to accurately identify an impulse sound with a small variation in sound pressure, and it is possible to identify the voice of the “sa” line, which was mistaken as an impulse in the conventional identification method, without being misidentified. become.

  Hereinafter, embodiments of an impulse identification device according to the present invention will be described with reference to the drawings. As shown in FIG. 1, an impulse identification device 10 (hereinafter referred to as “device 10”) is connected to a microphone 20 which is an acoustoelectric converter used for voice input and an amplifier 30 used to output an input signal. The amplifier 30 is used with a speaker 40 connected thereto. The apparatus 10 determines whether or not this input signal is an impulse sound because the voice of the speaker and ambient noise (steady killing sound and impulse sound) are mixed, and noise from the input signal. Only the component is removed and only the audio component is output to the amplifier 30.

  FIG. 2 is a functional block diagram showing an image of a circuit configuration provided in the microphone according to the present invention. The apparatus 10 includes an A / D 11 that converts a signal input from the microphone 20 into a digital signal, a fast Fourier transformer (FFT) 12 that converts the input signal converted into the digital signal into a frequency characteristic, and a harmonic in the frequency characteristic. A harmonic determination unit 13 that determines whether or not a wave component is included; a level determination unit 14 that determines a sound pressure level of an input signal; and a sound pressure level that does not include a harmonic component is higher than a predetermined sound pressure level. An average noise calculation unit 15 that holds the frequency characteristics of the small input signal and calculates an average value, a subtraction processing unit 16 that subtracts the frequency characteristic held by the average noise calculation unit 15 from the frequency characteristics of the input signal, and a subtraction process An inverse Fourier transformer (hereinafter referred to as “IFFT”) 17 that performs inverse Fourier transform on the result of the above, D / A that converts the signal subjected to inverse Fourier transform into an analog signal and outputs the analog signal to the amplifier 30 It is equipped with a 8. The FFT 12, the harmonic determination unit 13, the level determination unit 14, the average noise calculation unit 15, and the IFFT 17 constituting the device 10 are mounted using a DSP.

  Next, an impulse identification method using the apparatus 10 will be described with reference to the flowchart of FIG. In FIG. 2, each processing step is expressed as S301, S302,. First, the signal input from the microphone 20 is converted into a digital signal by the A / D 11 and then converted into frequency characteristics by the FFT 12 (S301). Next, the harmonic determination unit 13 determines whether a harmonic exists in the frequency characteristic (S302). When the harmonic determination unit 13 determines that there is a harmonic in the frequency characteristics of the input signal (Y in S302), the input signal is a voice, so the average noise held by the average noise calculation unit 15 is calculated. The average noise is subtracted from the frequency characteristics of the input signal (S306). Details of the average noise will be described later. The result of the subtraction is subjected to inverse Fourier transform in IFFT 12 (S307), and the changed digital signal is converted into an analog signal by D / A 18 and output to amplifier 30.

  When the harmonic determination unit 13 determines that no harmonic exists (N in S302), it is determined whether or not the gain at a predetermined frequency included in the frequency characteristic is equal to or higher than a predetermined level (S303). ). If the gain at the predetermined frequency is equal to or higher than the predetermined level (Y in S303), it is determined that the input sound is an impulse sound, the process of outputting the frequency characteristic is not performed, and the process returns to S301.

  When the harmonic determination unit 13 determines that there is no harmonic (N in S302) and determines that the gain at a predetermined frequency included in the frequency characteristic is not equal to or higher than a predetermined level (N in S303). ), It is determined that the frequency characteristic is noise, the average noise is calculated by the average noise calculation unit 15 using the frequency characteristic, and the average noise calculation unit 15 holds the calculated average noise (S305). Here, the average noise is obtained by calculating the time average of the frequency characteristics of the input signal when the input signal is other than sound and other than impulse sound. That is, the average value of ambient noise included in the input signal. By using the frequency characteristic of the average noise to perform subtraction processing on the frequency characteristic of the input signal determined to be speech, the ambient noise component included in the frequency characteristic determined to be speech can be removed. The frequency characteristics of the input signal with the noise canceled can be obtained. A clear audio output can be obtained by generating an output signal based on the frequency characteristic in which the noise is canceled.

  The details of the above impulse identification method will be described with reference to the waveform diagram of FIG. FIG. 4A shows an example of the frequency characteristic of the input signal. In FIG. 4A, the horizontal axis indicates the frequency, and the vertical axis indicates the gain (level) of each frequency. The frequency characteristic shown in FIG. 4A has a plurality of harmonics and has a certain gain or more. Therefore, the frequency characteristics shown in FIG. 4A can be recognized as frequency characteristics of vowels in particular among voices (human voices). The frequency characteristics shown in FIG. 4B can be recognized as ambient noise because there are no harmonics and the gain level is low. The frequency characteristic shown in FIG. 4C is a frequency characteristic of an impulse sound. Thus, the impulse sound has no higher harmonics and has a gain greater than a certain level. In this way, by detecting the peculiar characteristics of the impulse sound in the frequency characteristic, it is possible to properly identify the impulse sound, and further, it is possible to distinguish the impulse sound from the ambient noise, so that the noise is effectively prevented. Can be reduced.

  According to the above impulse sound identification method, even if the input signal is a voice, if it is a consonant, it can be reduced as noise without being misidentified as an impulse. By applying this method to mute, if it is an impulse sound, it will be muted, if it is a voice and consonant only, it will be processed as noise, and if other sounds are input, the noise will be reduced and output. Will be able to.

It is the schematic which shows the structure which connected the impulse identification based on this invention to the microphone. It is a block diagram which shows the example of a function structure of the microphone which implements the impulse identification concerning this invention. It is a flowchart which shows the example of a process of the impulse identification method of the said microphone. It is a graph which shows the frequency characteristic of human voice (a, b) and impulse sound (c).

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Microphone 12 Fast Fourier transformer 13 Harmonic determination part 14 Level determination part 15 Average noise calculating part 16 Subtraction process part 17 Inverse Fourier transformer

Claims (2)

An impulse identification device comprising a digital signal processor that determines whether or not the signal is an impulse by performing signal calculation processing on an input signal input from an acoustoelectric converter that converts ambient sound into an electric signal A Fourier transform means for transforming the input signal into a frequency characteristic;
First determination means for determining whether or not harmonics are included in the frequency characteristics;
An impulse identification device comprising: second determination means for determining whether or not a gain level of a predetermined frequency of the frequency characteristic is equal to or higher than a predetermined value. Fourier transform means for converting an input signal input from an acoustoelectric converter that converts ambient noise into an electrical signal into frequency characteristics, and first determination means for determining whether or not harmonics are included in the frequency characteristics And a second determination means for determining whether or not a gain level of a predetermined frequency of the frequency characteristic is equal to or higher than a predetermined value, and an impulse identification method using a microphone,
Converting the input signal into a frequency characteristic by the Fourier transform means;
A step of determining whether or not a harmonic is included in the frequency characteristic by the first determination means;
Determining whether or not a gain level of a predetermined frequency of the frequency characteristic determined by the second determination unit as being not included in the first determination unit is greater than or equal to a threshold value. Impulse identification method characterized by this.

Images