JPH0332799B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION Field of Industrial Application The present invention relates to a plosive classification device that can be used for speech recognition or for correcting the pronunciation of speech-impaired people.

Conventional structure and problems Conventionally, plosive sounds can be identified by detecting frequency components characteristic of plosive sounds from sound waves detected by a microphone using a filter bank or computational processing such as fast Fourier transform (FFT). However, the duration of plosives is short, about 10ms to 20ms, and there is no known method to reliably extract each phoneme of plosives.

Purpose of the Invention The purpose of the present invention is to detect the sounds accompanying speech, changes in oral pressure, oral airflow velocity, and opening and closing of the lips with respective detectors, and to detect the phonology of plosive sounds /p/, /
It is an object of the present invention to provide a plosive classification device that uses signal parameters characteristic of t/ and /k/ as standard parameters, identifies them by comparing them with input parameters, and reliably classifies plosives into each phoneme.

Structure of the Invention The present invention provides an oral pressure detector and voice detection device that can directly detect the increase in oral pressure that appears during speech by utilizing the relationship between oral pressure, voice, oral airflow, and lip opening/closing during speech. Microphone, mouth airflow detector that detects the flow of breath accompanying speech, upper lip,
A lip proximity detector that detects the proximity of the lower lip is used to extract the above-mentioned characteristic parameters associated with speech, and when plosives are uttered, the plosives are classified into each phoneme, /p/, /t/, /
Since this device directly detects the most basic parameters during speech, there are no false positives due to individual differences compared to conventional plosive detection methods, and reliable plosive classification is possible. .

DESCRIPTION OF EMBODIMENTS Hereinafter, embodiments of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram showing a plosive classification device in one embodiment of the present invention. In Fig. 1, 1 is a microphone that detects the sound during speech, and 2 is an oral pressure detector that detects changes in the atmospheric pressure in the oral cavity during speech, such as a small pressure detector such as a semiconductor strain gauge. configured. 3 is an oral air flow detector that detects the breath flow velocity associated with speech;
For example, it can be configured with a hot wire flowmeter or a condenser microphone. 4 is a lip proximity detector that detects the open/closed state of the lips; for example, by attaching a magnet to the upper lip and attaching a magnetic force sensing element such as a Hall element to the lower lip, the degree of proximity between the upper and lower lips can be detected. Reference numeral 5 denotes a voice rise detection circuit for detecting a rise from silence to sound in the voice output signal from the voice detection microphone 1, and is composed of an integrating circuit and a threshold circuit. 6 is a delay circuit that delays the output of the voice rise detection circuit 5 for a certain period of time; 7 is a characteristic parameter that stores signals from the oral pressure detector 2, oral airflow detector 3, and lip proximity detector 4 as characteristic parameters for a certain period of time. It is a memory circuit and is composed of an A-D conversion circuit and a memory circuit. 8 represents each plosive phoneme (/p/, /t/, /
This is a standard parameter storage circuit that stores feature parameters specific to the 1000p k/) in advance as standard parameters, and can be configured with a general memory circuit. 9
is triggered by the output of the delay circuit 6 which delays the signal from the voice rise circuit 5, and the characteristic parameters stored in the characteristic parameter storage circuit 7 and the plosive sounds stored in advance in the standard parameter storage circuit 7 are This is a phoneme classification circuit that compares phonemes with standard parameters and classifies each phoneme based on similarity, and is composed of a comparison circuit and the like.

FIG. 2 is a layout diagram showing the arrangement of signal detectors 1 to 4 in FIG. 1. In FIG. 2, the microphone 1 is placed in front of the mouth together with the oral airflow detector 3, and detects the flow velocity of the oral airflow associated with sound and speech, respectively. The intraoral pressure detector 2 is fixed to the hard palate in the oral cavity with an adhesive or the like, and detects the atmospheric pressure in the oral cavity. The lip proximity detector 4 measures the distance between the upper and lower lips by fixing a small magnet on the upper lip and a magnetic force-sensitive element such as a Hall element on the lower lip and detecting the intensity of magnetic lines of force.

According to experiments, the plosive sounds /p/, /t/, /k/
During speech, oral pressure, oral airflow, and lip position are different, and phonemes other than plosives (e.g., vowels, fricatives) are different, and plosives are determined by the above three characteristic parameters (oral pressure, oral airflow, and lip position). Classification of /p/, /t/, and /k/ is possible.

In other words, in general, plosive sounds completely close the oral cavity or vocal tract with the tongue or lips to store exhaled air, so when the intraoral pressure increases and reaches a certain limit, plosive sounds occur, producing impulse sounds and oral airflow. . The location where the closure is formed differs depending on the phoneme of the plosive. For example, when pronouncing /p/, the closure is made with the lips, as shown in Figure 3a, and when pronouncing /t/, the closure is created with the lips as shown in Figure 3b. A closure is created by the tongue and front teeth, as in When pronouncing /k/, a closure is formed deep inside the oral cavity, as shown in Figure 3c. In addition, phonemes other than plosives, such as fricatives, are produced by causing exhaled air to flow into a narrow space formed by the tongue and palate, causing turbulence, as shown in Figure 3d. Therefore, with the onset of speech, there is a relatively gradual increase in intraoral pressure and oral airflow due to the narrowing. When pronouncing vowels, the mouth is opened as shown in Figure 2, so there is no increase in intraoral pressure and no impulse air flow.

FIG. 4 shows the states of the three types of characteristic parameters at the time of each of the phonetic utterances in FIGS. 3a to 3d and FIG. 2, with reference to the time of voice rise. For example, in the plosive /p/ utterance, the third
As shown in Figure 4A, the lips are closed to hold the exhaled air, and when the lips are suddenly released, the voice is uttered at the same time, so the characteristic parameters before and after the rise of the voice are Closure and increase in intraoral pressure are observed, and at the same time as the onset of speech, a sharp pulse-like oral airflow occurs, and a rapid decrease in intraoral pressure and opening of the lips are observed. In the case of the plosive sound /t/, the closure peculiar to the plosive sound is /, as shown in Figure 3b.
Unlike the lips in p/, they are formed by the tip of the tongue and the upper front teeth, so as shown in Figure 4B, the state of the characteristic parameters before and after the rise of the voice depends on the intraoral pressure as in /p/. A rise and a sudden fall, as well as the generation of pulse-like oral airflow, are seen, but unlike the case of /p/, the lips are located far apart. In addition, in the case of /k/, a closure is formed by the tongue at the back of the oral cavity as shown in Fig. 3c, and a pulsed oral airflow is generated as shown in Fig. 4C. However, the rise and sudden drop in intraoral pressure before the rise of the voice seen in /p/ and /t/ and the closing of the lips seen in /p/ are not observed. In addition, for phonemes other than plosives, such as fricatives, as shown in Figure 3 d, the palate and tongue make a constriction, and as shown in Figure 4 D, the pressure in the mouth is relatively relaxed at the same time as the sound rises. An increase in air flow and the occurrence of oral airflow are seen. Since vowels are pronounced with the mouth wide open, there are no characteristic changes in oral pressure, oral airflow, or lip position before and after the onset of the sound, as shown in Figure 4E.

The present invention utilizes such a phenomenon to extract signals from the oral pressure detector 2, oral airflow detector 3, and lip proximity detector 4 as characteristic parameters, and extract the signals from the voice rise detection circuit 5 to a delay circuit. The characteristic parameters are stored in the characteristic parameter storage circuit 7 for a certain period of time until the trigger signal passed through step 6 is received. According to experiments, the increase in intraoral pressure and lip closure before the onset of speech as seen in Figures 4A and B last for 100ms to 200ms, depending on the phoneme, and the plosives after the onset of speech. Since the consonant part of a fricative is 200 to 300 ms, the feature parameter storage circuit 7 only needs to be able to hold data for 500 ms at a sampling interval of 10 ms, and old data can be deleted one after another. When a voice is produced, the voice is converted into an electrical signal by the microphone 1, and the voice rise detection circuit 5 detects the transition point from silence to sound, but the delay circuit 6 converts the voice into an electrical signal for a certain period of time (according to experiments). The longest fricative /∫/ is 200m~
After a delay of 300 ms, the signal is input to the phoneme classification circuit 9 as a trigger signal. When the trigger signal is input, the phoneme classification circuit 9 selects each phoneme/plosive sound stored in the standard parameter storage circuit 8 in advance.
The standard parameters of p/, /t/, /k/ are compared with the input feature parameters stored in the feature parameter storage circuit 7 based on the start of the voice, and the similarity is calculated. Classify sounds into each phoneme.

Effects of the Invention The present invention improves the sound quality and the changes in oral pressure associated with speech.
The velocity of the mouth airflow and the opening/closing state of the lips are detected by each detector, and outputs from the above detectors characteristic of each plosive phoneme (/p/, /t/, /k/) before and after the onset of the voice are detected. By storing in advance as a standard parameter and comparing the standard parameter with the input characteristic parameter when speech is input, it is possible to reliably classify plosives into each phoneme. Furthermore, even if a plosive occurs in a word, there is always silence before the plosive, so the present invention, which uses the rising point from silence to sound as a reference point for recognition, reliably classifies plosives into each phoneme. can do.

[Brief explanation of the drawing]

Fig. 1 is a block diagram of a plosive classification device according to an embodiment of the present invention, Fig. 2 is a layout diagram showing an embodiment of the mounting state of various detectors used in the present invention, and Figs. 3 a to d are FIG. 4 is a schematic diagram showing the articulation style during a typical phonological utterance, and is a correspondence diagram between each phoneme and the waveform of each feature parameter during a typical phonological utterance. 1...Microphone, 2...Oral pressure detector, 3
... Mouth airflow detector, 4... Lip proximity detector, 5...
...Audio rise detection circuit, 6...Delay circuit, 7...
. . . Feature parameter storage circuit, 8 . . . Standard parameter storage circuit, 9 . . . Phonological classification circuit.

Claims (1)

[Claims] 1. A voice detection microphone placed in front of the mouth, an oral pressure detector placed in the oral cavity, an oral airflow detector placed in front of the mouth to detect the oral airflow associated with speech, and the opening and closing of the lips. a lip proximity detector that detects a sound, a voice start-up detection circuit that detects a rise from silence to sound from an audio signal detected by the voice detection microphone, and a delay circuit that delays the signal of the voice start-up detection circuit. , a feature parameter storage circuit that stores feature parameters extracted by the oral pressure detector, the oral airflow detector, and the lip proximity detector for a certain period of time; and a standard parameter storage circuit that stores the feature parameters in advance;
Using a signal obtained by delaying the output signal of the speech rise detection circuit for a certain period of time as a trigger, the contents of the characteristic parameter storage circuit are compared with the contents of the standard parameter storage circuit for plosives, and each plosive is classified into phonemes. A plosive classification device comprising a circuit.