JPS6260720B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a nasal identification device that identifies nasal sounds only from words containing nasal consonants.

Conventionally, there is no known method for reliably detecting nasal sounds by analyzing audio signals.

However, the inventor of the present invention confirmed through experiments that the nasal interval can be detected using the three signals of nasal vibration, voice signal, and laryngeal vibration, and first used the three signals of nasal vibration, voice signal, and laryngeal vibration. We have provided a nasal sound detection device that detects only the nasal sound section reliably, easily, and in real time.

The nasal sound detection device will be described below with reference to the drawings.

FIG. 1 shows a block diagram of an embodiment of the nasal sound detection device provided above. In FIG. 1, reference numeral 1 denotes a nasal vibration detector for detecting vibrations of the nose, which measures the vibrations of the nose caused by speech and outputs an electrical signal, and is composed of, for example, a vibration pickup or the like. Reference numeral 2 denotes a nasal envelope circuit for extracting the envelope of the output signal of the nasal vibration detector 1, and is composed of, for example, a detection smoothing circuit. Reference numeral 3 denotes a microphone for detecting voice, which is composed of, for example, a low-noise close-talk type microphone. Reference numeral 4 denotes a high-pass filter circuit that removes signals of 25 to 35 [KHz] or less from the output signal of the microphone 3, that is, the audio signal, and is composed of, for example, an active filter. 5 is a high pass filter circuit 4
This is an audio envelope circuit that extracts the envelope of the output signal, and is composed of, for example, a detection and smoothing circuit. 6
This is a laryngeal vibration detector that detects the vibration of the larynx, which measures the vibration of the larynx caused by speech and outputs an electrical signal, and is composed of, for example, a vibration pickup. Reference numeral 7 denotes a laryngeal envelope circuit, which is composed of, for example, a detection smoothing circuit. 8 is a gate circuit that detects the output signal of the nasal envelope circuit 2 only during the generation period of the output signal of the laryngeal envelope circuit 7 and the non-generation period of the voice envelope circuit 5, and is, for example, an analog switch circuit, a digital gate circuit, etc. configured.

In the device configured as above, when a word containing a nasal sound is uttered, only the nasal sound section is processed by the gate circuit 8.
can be detected more easily. However, depending on the way of speaking, sometimes consonant intervals such as non-nasal sounds other than nasal sounds, such as voiced plosives with large nasal vibrations, may also be detected as nasal intervals from the output of the gate circuit 8. It has become clear that only nasal sounds can be specifically detected, which hinders their detection as nasal sounds. Therefore, it has become clear that it is difficult to detect nasal sounds from the output signal of the gate circuit 8, that is, from the nasal sound section.

According to experiments, the relationship between nasal vibration during the utterance of words containing nasal consonants and nasal vibration during the utterance of words containing non-nasal consonants such as voiced plosives with large nasal vibrations is as follows: Vibration magnitude (NC)
and the magnitude (nc) ratio of nasal vibration during the non-nasal consonant period,
The magnitude of nasal vibration in the vowel period following a nasal consonant (NV) and the magnitude of nasal vibration in the vowel period following a non-nasal consonant (nv)
Comparing the ratios, it was confirmed that the nasal vibration ratio in the subsequent vowel period was significantly larger than that in the consonant period.

NC/nc<NV/nv Therefore, rather than classifying the nasal interval signal into nasal and non-nasal sounds based on the magnitude of nasal vibration during the consonant period, the nasal interval signal is classified into nasal sounds based on the magnitude of nasal vibration during the vowel period. It was confirmed that it is reliable to separate and extract non-nasal sounds. The present invention utilizes this.

An object of the present invention is to provide a nasal sound identification device that reliably and easily identifies nasal sound signals regardless of the manner of speech.

Embodiments of the present invention will be described below with reference to the drawings.

FIG. 2 shows a block diagram of a nasal sound identification device according to an embodiment of the present invention. In Fig. 2, 1 is a nasal vibration detector, 2 is a nasal envelope circuit, 3 is a microphone, 4 is a high-pass filter circuit, and 5 is a nasal vibration detector.
is a voice envelope circuit, 6 is a laryngeal vibration detector,
7 is the laryngeal envelope circuit, 8 is the gate circuit,
The above configuration is the same as that of the nasal sound detection device shown in FIG. Reference numeral 9 denotes a delay circuit that delays the output signal of the gate circuit 8, and is composed of, for example, a charge transfer element such as a BBD. 10 is an amplitude detection circuit for detecting the amplitude of the vowel period of the output signal of the nasal envelope circuit 2; for example, an analog switch circuit,
Consists of comparator circuits, etc. A processing circuit 11 outputs a nasal signal only when there is an output from the amplitude detection circuit 10 and an output from the delay circuit 9, and is composed of, for example, a voltage-controlled variable gain circuit or an AND circuit.

Figures 3 and 4 are non-nasal sounds with large nasal vibrations, ``ide'', which are produced by the nasal sound identification device shown in Figure 2. )” is shown, and the signal waveforms of each component are shown below.
The operation of the nasal sound identification device will be explained with reference to FIG.

In FIGS. 3 and 4, waveforms A and a show the audio signals detected by the microphone 3, and waveforms B and b show the output signals of the high-pass filter circuit 4. In addition, waveforms C and c are from the nasal vibration detector 1.
Waveforms D and d show signals of laryngeal vibration detected by the laryngeal vibration detector 6. Waveforms E and e are output signals of the audio envelope circuit 5, and are obtained by extracting the envelopes of the output signals B and b of the high-pass filter circuit 4. Waveforms F and f are output signals of the nasal envelope circuit 2, and output signals C and F of the nasal vibration detector 1 are
This is an extracted envelope of c. Waveforms G and g are output signals of the laryngeal envelope circuit 7, and are obtained by extracting the envelopes of the output signals D and d of the laryngeal vibration detector 6. Waveforms H and h indicate speech gate signals that detect the non-occurrence period of speech envelope signals E and e, and waveforms I and ii indicate laryngeal gate signals that detect the occurrence period of laryngeal vibration envelope signals G and g. These voice gate signals H, h and laryngeal gate signals G, g are generated inside the gate circuit 8. Waveforms J and j indicate the output signals of the gate circuit 8, and envelope signals F and f of nasal vibration are generated only during the generation period of the laryngeal gate signals I and i and the voice gate signals H and h.
was detected. That is, the gate circuit 8 is triggered by the 0 level of the speech envelope signals E, e, and generates the speech gate signals H, h representing the non-occurrence period of the speech envelope signals E, e, as well as the laryngeal envelope signals G, g. is triggered at the 0 level of , and generates laryngeal gate signals I,i representing the period of occurrence of the laryngeal envelope signals G,g. Furthermore, the gate circuit 8 outputs the nasal envelope signals F, f only during the generation period of the laryngeal gate signals I, i and only during the generation period of the voice gate signals H, h. Therefore, a nasal segment signal J of a nasal segment T is detected from the word "ini", and a nasal segment signal j of a nasal segment t is detected from the word "ide". Therefore, as mentioned above, depending on the way of speaking,
The output signal j of the gate circuit 8 for the word "ide" containing a non-nasal sound with large nasal vibrations is similar to the output signal J of the gate circuit 8 for the word "ini" containing a nasal sound, that is, the nasal interval signal. It becomes difficult to distinguish between nasal sounds and non-nasal sounds using the output signals J, j.

On the other hand, waveforms K and k indicate the output signals of the delay circuit 9, and the output signals J and j of the gate circuit 8 are delayed by a delay time α
(The dashed line shows the signal before delay). Here, the number of stages and the clock frequency of the charge transfer elements constituting the delay circuit 9 are such that the output signal K, k of the gate circuit 8 after the delay is the nose envelope circuit 2.
is set to provide a delay time α that is located within the vowel period following the nasal consonant or non-nasal consonant of the output signal. Waveforms L and l are amplitude detection circuit 1
0 output signal, and becomes Hi when the output signal of the nasal envelope circuit 2 is at a level equal to or higher than a predetermined threshold value υ within the output signal generation period of the audio envelope circuit 5.
This is what triggered the level. Here, the comparator circuit constituting the amplitude detection circuit 10 is used when the output signal of the nasal envelope circuit 2 is below the nasal vibration level when a nasal consonant or a non-nasal consonant is uttered, and when a non-nasal consonant is uttered. It is set to give a threshold value υ such that the nasal vibration level is higher than or equal to the nasal vibration level. Therefore, since the nasal vibration in the vowel period of a word containing a nasal sound is larger than the nasal vibration in the vowel period of a word including a non-nasal sound, the output signals of the vibration detection circuit 10 have waveforms L and l, respectively. Therefore, if the delayed nasal interval signals K, k are processed by the processing circuit 11 which outputs a nasal signal from the output signal of the delay circuit 9 when there is an output signal of the amplitude detection circuit 10, the waveform M as shown in FIG. Nasal signals can be detected only in words that include nasal sounds, and in words that include non-nasal sounds with large nasal vibrations, no nasal signal is detected as shown in waveform m, and misrecognition can be prevented. .

As described above, the nasal sound identification device of the present invention includes a nasal vibration detector that detects nasal vibrations, a microphone that detects voice, a high-pass filter circuit that filters the output signal of the microphone, and a nasal vibration detector that detects vibrations of the larynx. A gate circuit that outputs the output of the nasal vibration detector as a nasal interval during a period when the laryngeal vibration detector and the laryngeal vibration detector are detecting vibration and there is no output from the high-pass filter circuit, and the output of the high-pass filter circuit. An amplitude detection circuit detects the magnitude of the output of the nasal vibration detector for a certain period, a delay circuit delays the output of the gate circuit, and the output of the delay circuit is controlled according to the magnitude of the output of the amplitude detection circuit. Because it is configured with a processing circuit, it is possible to detect nasal sounds only from words that contain nasal consonants, without detecting nasal sounds from words that contain non-nasal consonants, regardless of how they are uttered, and its industrial value is great. There is something.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of a nasal sound detection device previously proposed by the present inventor, FIG. 2 is a block diagram of a nasal sound identification device that is an embodiment of the present invention, and FIGS.
Figures A to W are signal waveform diagrams of each component of the device. DESCRIPTION OF SYMBOLS 1... Nasal vibration detector, 2... Nasal envelope circuit, 3... Microphone, 4... High pass filter circuit, 5... Audio envelope circuit, 6... Laryngeal vibration detector, 7... Laryngeal envelope circuit, 8 ……
Gate circuit, 9...delay circuit, 10...amplitude detection circuit, 11...processing circuit.

Claims (1)

[Claims] 1. A nasal vibration detector that detects nasal vibrations, a microphone that detects voice, a high-pass filter circuit that filters the output signal of the microphone, a laryngeal vibration detector that detects laryngeal vibrations, and the laryngeal vibration detector. a gate circuit that outputs the output of the nasal vibration detector as a nasal section during a period in which the device is detecting vibrations and there is no output from the high-pass filter circuit; an amplitude detection circuit that detects the magnitude of the output of the detector;
A nasal sound identification device comprising: a delay circuit that delays the output of the gate circuit; and a processing circuit that controls the output of the delay circuit according to the magnitude of the output of the amplitude detection circuit.