Method and sound isolation device for separating and collecting nasal and oral sounds in non-nasalized vowels

Abstract

The application discloses a method and a sound insulation device for separating and collecting nasal sound and oral sound in non-nasalized vowel, and belongs to the field of speech signal processing, and aims to solve the problem that the traditional separation method is not applicable to audio collection. The sound insulation device comprises a nasal sound muffler and an oral sound muffler; the nasal sound muffler is a hanging ear type nasal air filter, the exhaust hole below the nasal mask of the hanging ear type nasal air filter and the outlet hole connected with the hanging rope on the side are closed by plastic by means of glue; the hanging ear type nasal air filter is punched above the nasal mask, and a medical bandage is placed in the nasal mask; the oral sound muffler is composed of a main mask, a long tube and a sound insulation box, the main mask is in the shape of a funnel, and the lower end of the main mask is in the shape of a circular tube; the main mask is surrounded by super clay. The sound insulation device is light in weight, easy to obtain and convenient to use; and the method can record relatively pure sounds transmitted from the oral cavity and sounds transmitted from the nasal cavity respectively.

Description

Technical Field

[0001] This invention relates to the field of speech signal processing, specifically to a method and a soundproofing device for separating and acquiring nasal and accent sounds in non-nasalized vowels. Background Technology

[0002] Sound is produced by the vibration of the glottis, modulated by the vocal tract, and finally radiated into space through the lips and nostrils. Speech contains rich information about the speaker's characteristics, and the technology for extracting this information has wide applications in finance, criminal investigation, and national defense. The personalized characteristics of speech are determined by the structure of the articulatory organs. From bottom to top, the articulatory organs include the glottis, larynx, pharynx, oral cavity, palate, nasal cavity, and lips. The oral cavity and nasal cavity are important resonating organs, playing an indispensable role in sound modulation. The oral cavity adjusts the shape of the oral passages by changing the position of the tongue, producing different syllables. The nasal cavity has a stable internal structure, and its morphology varies significantly among different speakers, containing rich information for distinguishing speaker identity. Studying the influence of nasal resonance on speech characteristics helps enrich research on the personalized characteristics of speakers.

[0003] The nasal cavity has a complex structure, consisting of two asymmetrical nasal passages connected to four pairs of paranasal sinuses: the frontal sinuses, maxillary sinuses, sphenoid sinuses, and ethmoid sinuses. Each paranasal sinus is a narrow-atlet cavity with a large internal surface area, resembling a Helmholtz resonator. Their acoustic function is to absorb certain frequencies of sound in nasal tones. The four pairs of sinuses vary considerably in size, absorbing different frequencies of sound. The asymmetry of the nasal passages also causes certain frequencies to be either elevated or lowered.

[0004] The nasal cavity influences pronunciation primarily in two ways. The first occurs during the articulation of nasal and nasalized vowels, where the raising and lowering of the soft palate controls the opening and closing of the velar orifice, a process known as oropharyngeal coupling. During the articulation of ordinary vowels, the soft palate typically rises to meet the palatal wall, closing the passage from the oral cavity to the nasal cavity. Nasalized vowels, distinct from ordinary vowels in Mandarin, refer to vowels preceding or following nasal consonants. Influenced by these consonants, the velar orifice is not fully closed, and the nasal cavity is considered a branch of the vocal tract. The second method involves sound transmission from the oral cavity to the nasal cavity via the vibration of the soft palate, a process distinguished from the first, known as trans-soft palate coupling. Trans-soft palate coupling is the main way the nasal cavity participates in the articulation of non-nasalized vowels and voiced stops. Sound transmitted to the nasal cavity radiates into space through the nostrils, interacting with lip-radiated sounds to form speech.

[0005] Experimental studies on the effects of soft palate coupling on speech acoustic features are scarce, and techniques for measuring nasal resonance independently are limited. Calculating the spectrum of nasal resonance using MRI images of the speaker to simulate nasal cavity morphology is a relatively accurate technique. However, the internal morphology of the nasal cavity is complex, and limitations in medical equipment make it difficult to achieve the required accuracy, or the cost is very high.

[0006] To study the effect of soft palate coupling on speech features, it is necessary to record the speaker's lip-radiated and nasal-radiated sounds separately. This requires using sound-insulating devices to separate the oral and nasal cavities and collect their audio frequencies. Traditional separation methods include using oronasal airflow meters and sound-insulating panels. Oronasal airflow meters are effective tools for detecting airflow and analyzing aerodynamic data of speech, but are not suitable for audio recording. Effectively preventing the mixing of lip-radiated and nasal-radiated sounds would require a large, thin sound-insulating panel. These materials are difficult to obtain and waste space, making them inconvenient to use. Neither method is ideal for sound insulation. Summary of the Invention

[0007] To address the aforementioned technical shortcomings, the purpose of this invention is to provide a method and a soundproofing device for separating and capturing nasal and oral sounds in non-nasalized vowels. This method can record relatively pure sounds transmitted from the oral cavity and sounds transmitted from the nasal cavity, respectively.

[0008] To solve the above-mentioned technical problems, the present invention adopts the following technical solution: This invention provides a soundproofing device, characterized in that it includes a nasal muffler and an oral muffler; the main body of the nasal muffler is an ear-hook nasal air filter, and the exhaust port below the ear-hook nasal air filter and the outlet port for connecting the hanging rope on the side are sealed with plastic by glue; the ear-hook nasal air filter has a hole at the top and a medical bandage is placed inside the nasal filter. The accent silencer consists of three parts: a main mask, a long tube, and a soundproof box. The main mask is funnel-shaped, and the lower end of the main mask is a circular tubular structure. The main mask is surrounded by super clay. One end of the long tube is glued to the lower end of the main mask, and a heat shrink tube is tightly wrapped around the long tube. The inner wall of the soundproof box is covered with a rubber sheet, and a sponge pad is attached to the rubber sheet. There are circular holes on both sides of the soundproof box, with one circular hole slightly higher and the other slightly lower. The other end of the long tube is inserted into the circular hole on one side of the soundproof box, and a cone-shaped sparse material is placed inside the other end of the long tube. A plastic round tube is inserted into the circular hole on the other side of the soundproof box.

[0009] Furthermore, the cone-shaped sparse material built into the other end of the long tube is made of wood or bamboo sticks.

[0010] This invention also relates to a method for separating and collecting nasal and accent sounds in non-nasalized vowels, characterized by comprising the following steps: Step 1: Wear a nasal muffler to eliminate interference from nasal radiation and record pure oral sounds. Step 2: Position the main body of the accent silencer mask towards your mouth, using the accent silencer to eliminate interference from lip-radiated sounds and record a relatively pure sound transmitted through the nasal cavity. Step 3: Record the dataset: Design the corpus, i.e., the audio text recorded by the speaker; select subjects and determine the number of participants; Three types of speech data were recorded: natural speech, oral and nasal sounds; natural speech, also known as oral-nasal mixed speech, was recorded normally by the subject wearing a microphone; when recording oral sounds, the subject wore a nasal muffler; when recording nasal sounds, the subject's mouth was placed in front of the nasal muffler. Step 4: Analyze the voice data collected in Step 3 to verify the effectiveness of the sound insulation device from three aspects; First, analyze the waveform of the speech and compare whether the loudness of the three types of speech data conforms to the pattern; the loudness is expected to be from largest to smallest as follows: natural speech, oral cavity transmitted sound, and nasal cavity transmitted sound; then, calculate the three audio amplitudes under the same text content to obtain the sound transmission loss. Finally, compare the waveforms of open and closed vowels recorded by the same speaker; compared to open vowels, closed vowels have a stronger coupling effect through the soft palate and should have a stronger nasal radiation sound; the typical open vowel is / a / , and the closed vowel is / i / .

[0011] The beneficial effects are: the materials used in this device are lightweight and readily available, and it is easy to use; it can record relatively pure sounds transmitted from the oral cavity and the nasal cavity respectively. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 This is a schematic diagram of the use of a nasal muffler provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the use of an accent silencer provided in an embodiment of the present invention; Figure 3 Three speech waveform diagrams of the speaker's ZY vowel / a / provided in embodiments of the present invention; Figure 4The waveform diagrams of the speaker's ZY vowels / a / and / i / are provided in the embodiments of the present invention.

[0014] Explanation of reference numerals in the attached diagram: 1. Nasal muffler; 2. Microphone; 3. Main mask; 4. Non-reflective tube; 5. Soundproof box. Detailed Implementation

[0015] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0016] Example 1 like Figures 1-2 As shown, a method for separating and collecting nasal and accent sounds in non-nasalized vowels includes the following steps: Step 1: Making a Nasal Voice Silencer 1. After comparing products from several manufacturers, we selected the ear-hook nasal air filter produced by Haze Star. This nasal mask has good sealing properties and fits snugly against the speaker's face. The ear-hook design makes it easy to wear. To maximize the reduction of nasal sound transmission while maintaining constant air pressure inside and outside the nasal cavity, we completely sealed the exhaust vent at the bottom of the nasal mask and the outlet for the side hanging strap using special plastic adhesive, and then drilled a new hole at the top of the nasal mask. Finally, we placed a medical bandage inside the nasal mask.

[0017] Step Two: Constructing the Accident Muffler. This soundproofing tool consists of three parts: a mask, a long tube, and a soundproof box 5. The mask is funnel-shaped, oval at the top and circular at the bottom, with the cross-section narrowing from top to bottom. Cut the front end of a mineral water bottle, leaving the opening intact, and surround the bottle with super clay to reduce sound leakage. The oval front end of the mask is of moderate length and width, fitting snugly around an adult's lips while exposing the nostrils. The circular rear end connects to a 1m long aluminum tube with an inner diameter of 3cm, secured with plastic adhesive. The entire aluminum tube is tightly wrapped with heat-shrink tubing. One end of the tube is connected to the mask, and the other end is inserted into the soundproof box 5. The end inserted into the soundproof box 5 contains a cone-shaped sparse material, made using wooden or bamboo skewers. The soundproof box 5 is made of wood, 40cm long, 30cm wide, and 16cm high. A 5mm thick rubber sheet is pasted inside the wooden board, and a 3cm thick sponge pad is pasted on the rubber sheet. Drill circular holes with a diameter of 3.5cm on both sides of the soundproof enclosure 5, one hole slightly higher and the other slightly lower. During recording, insert a long aluminum tube into one hole on each side of the soundproof enclosure 5, and a plastic tube with a diameter of 3.4cm into the other hole. The insertion distance is 10cm for each hole.

[0018] Step 3: Record the dataset. The corpus consists of two parts: Chinese vowels and Chinese short sentences. Vowels include / a / , / o / , / e / , / i / , / u / , and / v / . The Chinese short sentence is "I also mistakenly thought the meaning was meaningless." One person, referred to as ZY, recorded the data. Each vowel was repeated three times during recording, and the short sentence was also repeated three times. The recording sound card was a Yamaha AG03, and microphone 2 was an Audio-Technica AT9904. The audio sampling frequency was 16kHz, the sampling bit depth was 16, the number of channels was single-channel, and the audio format was WAV. When recording natural speech, microphone 2 was placed 5cm away from the lips. When recording lip-radiated sounds, a modified nose mask was worn, and microphone 2 was placed in the same position. When recording nasal-radiated sounds, the lips were placed at one end of the accent mask, maintaining a seamless fit around the lips, and microphone 2 was placed 2cm away from the nostrils.

[0019] Step 4: Analyze the data. Use the mathematical software MATLAB to analyze the audio data obtained in Step 3. Plot a waveform graph, with time on the horizontal axis and loudness on the vertical axis. Plot a spectrum graph, with frequency on the horizontal axis and amplitude on the vertical axis. To depict the characteristic trend of the spectrum graph, calculate the spectral envelope using cepstral analysis and plot it on the spectrum graph. The vocal tract shape of vowels is static, and their spectrum graph can be a short-time Fourier spectrum graph, selecting any frame from a stationary signal segment. The vocal tract shape corresponding to short phrases is dynamic, and a long-time average Fourier spectrum graph is suitable for the spectral analysis of short phrases, that is, the spectrum of each frame is added together and averaged. Specific parameters are shown in Table 1.

[0020] Table 1. Short-time speech spectrogram parameters Frame length 100ms Frame shift 50ms Fourier points 4096 Window function Hanning Window Data analysis results as follows Figures 3-4 As shown in Table 2, a direct comparison of the waveforms of the vowel / a / reveals that normal speech has the highest loudness, while the loudness transmitted through the nasal cavity is the lowest. The sound transmission loss at 1kHz is 34dB. Comparing the waveforms of the vowels / a / and / i / , the normally recorded vowel / a / has a stronger loudness and a weaker nasal resonance. The experimental results are consistent with the expectations of step four in the design scheme.

[0021] Table 2. Sound transmission loss of speaker ZY via soft palate coupling. nasal -50 -53 -54 -55 -52 Normal speech -17 -19 -21 -27 -26 Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A sound insulation device, characterized in that, It includes a nasal muffler and an oral muffler; the main body of the nasal muffler is an ear-hook nasal air filter, and the exhaust port below the ear-hook nasal air filter and the outlet port on the side for connecting the hanging rope are sealed with plastic glue. The ear-hook nasal air filter has a hole at the top and a medical bandage is placed inside the nasal mask; the accent silencer consists of three parts: the main mask, the long tube, and the soundproof box. The main mask is funnel-shaped and the lower end of the main mask is a circular tubular structure; the main mask is surrounded by super clay. One end of the long tube is glued to the lower end of the main mask, and heat shrink tubing is tightly wrapped around the long tube. The inner wall of the soundproof box is lined with a rubber sheet, and a sponge pad is attached to the rubber sheet; circular holes are provided on both sides of the soundproof box, with one circular hole slightly higher and the other slightly lower. The other end of the long tube is inserted into a circular hole on one side of the soundproof box, and a cone-shaped sparse material is placed inside the other end of the long tube; a plastic round tube is inserted into the circular hole on the other side of the soundproof box.

2. The sound insulation device according to claim 1, characterized in that, The cone-shaped sparse material inside the other end of the long tube is made of wooden bamboo sticks.

3. A method for separating and collecting nasal and accent sounds in non-nasalized vowels using the sound insulation device described in claim 1, characterized in that, Includes the following steps: Step 1: Wear a nasal muffler to eliminate interference from nasal radiation and record pure oral sounds. Step 2: Position the main body of the accent silencer mask towards your mouth, using the accent silencer to eliminate interference from lip-radiated sounds and record a relatively pure sound transmitted through the nasal cavity. Step 3: Record the dataset: Design the corpus, i.e., the audio text recorded by the speaker; select subjects and determine the number of participants; Three types of speech data were recorded: natural speech, oral and nasal sounds; natural speech, also known as oral-nasal mixed speech, was recorded normally by the subject wearing a microphone; when recording oral sounds, the subject wore a nasal muffler; when recording nasal sounds, the subject's mouth was placed in front of the nasal muffler. Step 4: Analyze the voice data collected in Step 3 to verify the effectiveness of the sound insulation device from three aspects; First, analyze the waveform of the speech and compare whether the loudness of the three types of speech data conforms to the pattern; the expected loudness from largest to smallest is: natural speech, sound transmitted through the oral cavity, and sound transmitted through the nasal cavity. Then, the three audio amplitudes under the same text content are calculated to obtain the sound transmission loss; Finally, compare the waveforms of open and closed vowels recorded by the same speaker. Compared to open vowels, closed vowels have a stronger trans-palm coupling effect and should have a stronger nasal radiation. The typical open vowel is / a / , and the closed vowel is / i / .

Images