KR20110005669A - Signal processing method of digital hearing aid - Google Patents

Abstract

PURPOSE: A method for processing the signal of a digital hearing aid is provided to automatically erase specific peripheral noises of a narrow frequency band using a spectrum noise erasing algorithm. CONSTITUTION: In a frame, a digital signal is processed with respect to 128 input signals. A final voice signal is transferred to an output buffer. A programming algorithm is implemented in order to synchronize 32 output buffers to a receiver in 0.0625msec of sampling time. A spectrum amplitude modulation signal is processed after a fast Fourier transform is implemented. A fast Fourier reverse transform algorithm is implemented to calculate output voice data. A spectrum amplitude modulation algorithm is implemented to mainly concentrate conversation voice.

Description

SIGNAL PROCESSING METHOD OF DIGITAL HEARING AID

BACKGROUND OF THE INVENTION 1. Field of the Invention [0001] The present invention relates to a hearing aid-type digital hearing aid. In particular, within the hearing threshold dynamic range of a hearing impaired person, the output sound pressure of the digital hearing aid is different for each frequency channel, and automatically removes acoustic feedback that may occur in wearing the hearing aid. In addition, it is possible to automatically cancel specific ambient noise in a narrow frequency band, and to change the tone according to the wearer's preference, as well as a digital hearing aid signal processing method that can implement a multipurpose hearing aid. It is about.

The digital signal processing method of the digital IC chip embedded in the ear insertion type hearing aids developed or marketed so far is to perform the methods for the compression of the audio signal, the removal of the acoustic feedback, the noise cancellation of the surrounding environment, and the tonal changes. As a result, the amount of computation of the central processing unit (CPU) in the digital IC chip increases, resulting in high power consumption and complicated execution program algorithms.

The present invention has been made in view of the above problems, and an object of the present invention is that the output sound pressure of the digital hearing aid is differently generated for each frequency channel within the hearing threshold dynamic range of the hearing impaired, and may occur in wearing a hearing aid. The present invention provides a signal processing method of a digital hearing aid.

Another object of the present invention is to provide a signal processing method of a digital hearing aid that can automatically remove acoustic feedback.

It is still another object of the present invention to provide a signal processing method of a digital hearing aid that can automatically cancel specific ambient noise of a narrow frequency band.

Still another object of the present invention is to provide a signal processing method of a digital hearing aid that can change a tone according to a wearer's taste.

Another object of the present invention is to provide a signal processing method of a digital hearing aid that can implement a multi-purpose performance hearing aid.

The present invention has been made to achieve the above object, and the signal processing method of the digital hearing aid of the present invention processes digital signals for 128 input signals per frame, but some signal dropouts are caused due to signal processing time by the CPU. In order not to occur, the 32 final voice signals that were input first are moved out of the last 128 frame data after all spectrum modulation algorithm signal processing is completed during the 0.0625msec sampling time as soon as 32 new voice signals are input through the input buffer. Performing a programming algorithm for synchronizing the 32 output buffers with a receiver at 0.0625 msec sampling time and outputting the same; After performing the programming algorithm, after converting only 65 complex data into log unit (dB HL) after fast free-switching, the spectrum amplitude modulation signal processing is performed, and then again converted from log unit to 128 complex units of expected value. Performing a fast free inverse transform algorithm that computes 128 completed final output speech data with a fast free inverse transform; If the spectrum of the voice report continuously input for 200 msec after the fast free inverse transform algorithm is smaller than the consonant threshold level curve of the conversation sound, it is regarded as ambient noise and the attenuation is not exposed to the amplified ambient noise by attenuation rather than amplification. And performing a spectral amplitude modulation algorithm so as to concentrate only on speech-oriented voices.

According to the signal processing method of the digital hearing aid according to the present invention, within the hearing threshold dynamic range of the hearing impaired, the output sound pressure of the digital hearing aid is generated differently for each frequency channel, and the signal processing of the digital hearing aid may occur when wearing a hearing aid. To provide a way.

Another object of the present invention is to provide a signal processing method of a digital hearing aid that can automatically remove acoustic feedback.

It is still another object of the present invention to provide a signal processing method of a digital hearing aid that can automatically cancel specific ambient noise of a narrow frequency band.

Still another object of the present invention is to provide a signal processing method of a digital hearing aid that can change a tone according to a wearer's taste.

Another object of the present invention is to provide a signal processing method of a digital hearing aid that can implement a multi-purpose digital hearing aid.

1 is a perspective view of an ear insertion type hearing aid to which the present invention is applied, and shows a typical external structure.
2 is a conceptual diagram illustrating a digital signal converted by an analog-to-digital converter of a digital IC chip according to an embodiment of the present invention.
3 is a conceptual diagram illustrating a series of memory buffer spaces (1 to 128 address addresses) including 13 bits per byte according to an embodiment of the present invention.
4 is a conceptual diagram illustrating an operation of calculating 128 expected complex data from 65 complex data according to an embodiment of the present invention.
5 is a conceptual diagram illustrating a data flow for digital signal processing in a first to fourth memory buffer space according to an embodiment of the present invention.
FIG. 6 is a conceptual diagram illustrating an operation operation of moving new data by 32 data while continuously performing digital signal processing with 128 memory buffers according to an embodiment of the present invention.
7 is a conceptual diagram illustrating an operation operation of performing a spectrum modulation algorithm and a fast free inverse transform after a fast free transform according to an embodiment of the present invention.
8 is a conceptual diagram showing the hearing test result table of the hearing loss according to an embodiment of the present invention.
9 is a conceptual view showing threshold level curves and a voice signal spectrum simultaneously by a hearing test according to an embodiment of the present invention.
FIG. 10 is a conceptual diagram illustrating the initial, neutral, and final characteristics of waveforms of a voice signal according to an embodiment of the present invention.
FIG. 11 is a conceptual diagram illustrating how to amplify or attenuate a voice signal in a frequency band according to the hearing condition of a hearing impaired person according to an embodiment of the present invention.
12 is a conceptual diagram showing the degree of matching the dynamic range of the dialog sound to the hearing loss range of the hearing impaired according to an embodiment of the present invention.
FIG. 13 is a conceptual diagram illustrating a dynamic range of a hearing test result table and a dialogue sound of a hearing impaired person according to an embodiment of the present invention together on a spectrum.
14 is a conceptual diagram illustrating an overall operation of a dynamic range compression algorithm among spectral amplitude modulation algorithms according to an embodiment of the present invention.
15 is a conceptual diagram illustrating an amplitude spectrum of speech according to an embodiment of the present invention.
16 is a conceptual diagram illustrating an average value of a voice amplitude spectrum and comparing the instantaneous amplitude spectrum according to an embodiment of the present invention.
FIG. 17 is a conceptual diagram illustrating an amplitude spectrum of an entire signal sound input together with voice when acoustic feedback or narrowband noise occurs according to an embodiment of the present invention.
FIG. 18 is a conceptual diagram illustrating an average value of amplitude spectrums when a voice amplitude spectrum is rapidly changed due to acoustic feedback or narrowband noise according to an embodiment of the present invention.
19 is a conceptual diagram illustrating a corresponding digital signal processing operation operation when acoustic feedback or narrowband noise is generated according to an embodiment of the present invention.

Hereinafter, a signal processing method of a digital hearing aid according to an embodiment of the present invention will be described in detail with reference to the accompanying drawings.

A signal processing method of a digital hearing aid according to an embodiment of the present invention will be described with reference to the drawings illustrated in FIGS. 1 to 19.

The digital signal converted by the analog-to-digital conversion module of the digital IC chip embedded in the earshell 10 of the digital hearing aid is continuously input voice signal strength with respect to the time axis (X axis) at regular time intervals (sampling time). It is denoted by the axis (Y axis).

The analog-digital converted voice signal data is collected at regular sampling time intervals and continuously stored in the internal memory of the digital IC chip. For example, each memory byte is stored in 13 bit units at 1 to 128 memory addresses. Separately input and output digital binary values of 0 or 1. A 13-bit binary number, in terms of decimal logarithmic (dB) scale, has a logarithmic scale range of 20 x log ₁₀ (2 ¹³ ), that is, about 78 dB. The hearing threshold of a hearing impaired person who needs a digital hearing aid increases from the normal threshold of about 25 dB, so that 25 dB to 78 dB plus 25 dB becomes the input / output dynamic range of the digital hearing aid.

When digital signal processing is performed while digitally inputting and continuously inputting binary audio signal data to and from 128-byte memory space, a sampling frequency of 0.0625 msec is assumed when the sampling frequency is 16 kHz. When 128 audio data are input and stored sequentially, the audio signal is collected for 8msec time multiplied by 128 times the sampling time of 0.0625msec, which is the inverse of 8msec, that is, the length of a sinusoidal signal cycle of 125Hz. This means that the minimum number of data needed to correct hearing thresholds for deaf people is 128. Performing a fast free conversion from 128 time domain real data, x (n), calculates 128 frequency domain complex data, X (n). N = 1 to 128 here. Because 128 complex data are expected to be 64 data, which is half of the total, only 65 data can be stored without having to store all of them. In order to perform the inverse fast free transform from the 65 frequency domain complex data calculated by the fast free transform, the remaining 63 (66-128) expected complex data from the 65 complex data immediately before the inverse transform are performed. Calculate first.

이다. 마찬가지로 X(67)=X(63)^* 이며 이와 같은 방식으로 모든 X(68),...,X(128) 에 대해 동일하게 기대칭 복소수 데이터들을 계산한다.For example, X (66) = X (64) ^* . For example, if X (64) = 0.5 + j2.4, then X (66) = 0.5-j2.4. Where j is

to be. Similarly, X (67) = X (63) ^* and in this way we compute equally expected complex data for all X (68), ..., X (128).

For example, there is a physical limitation in the computational capability of a digital IC chip to perform all the digital signal processing operations required to realize the performance of a digital hearing aid during a sampling time of 0.0625 msec. Therefore, when the operation time required for the performance of the implementation of _a digital hearing aid T d in view of the computational power of the digital IC chip, e.g., _a digital T for 32 times the sampling time by dividing the sampling time by 2msec 0.0625msec Speaking 2msec The IC chip must perform all the operations necessary to implement the performance of the digital hearing aid. In this case, _a part of the voice signal which is continuously input continuously during the operation time of T _a , or 2 msec digital IC chip, is cut off. Therefore, in order to fundamentally solve such a disconnection phenomenon, the present invention has devised a new calculation method.

1. The proposed technique proposes a programming algorithm in the following order to execute all the operations required to implement the performance of a digital hearing aid for 2 msec with 128 speech signal time domain data.

The primary memory buffer space (input buffer) is a place to collect and store the input voice signal first, and the secondary memory buffer space (preprocessing buffer) is to sequentially store 32 binary data of the primary memory buffer space into four groups. The tertiary memory buffer space (post-processing buffer) is a place to temporarily store and store the result of digital signal processing with 128 binary data of the secondary memory buffer space, and the fourth memory buffer space. (Output buffer) is a place to wait to output to the external receiver after moving the 32 binary data stored first from the tertiary memory buffer space.

1-1) The audio signal data continuously input by the FIFO (First In First Out) method is sequentially stored (1,2,3, .., 32) one by one at a sampling time interval in 32 primary memory buffer spaces. .

1-2) With 128 binary data (1, 2, ... 128) contained in 128 secondary memory buffer spaces for 2 msec time, the above 1-1) is executed. After parallel execution of all the operations required for performance, the final result is moved to tertiary memory buffer space. The 128 secondary and tertiary memory buffer spaces are divided into four 32 memory buffer spaces (G1, G2, G3, and G4).

1-3) When all 32 audio signal data are newly input into the primary memory buffer space in 1-1) and 1-2) above, within the 0.0625 msec sampling time, first, the tertiary memory buffer space 32 data in the G4 group of the data are moved in parallel to 32 data positions in the 4th memory buffer space, and 96 data in the G1, G2, and G3 groups in the secondary memory buffer space are G2, G3, Shift to 96 data spaces in the G4 group, and the primary memory buffer space moves parallel to 32 data locations in the G1 group of secondary memory buffer space. In the central processing of the digital IC chip, the memory data movement speed is much higher than that of arithmetic operations, so the time required for two 32 memory buffer data parallel movements and 96 memory buffer data shift movements is very small. It is possible to implement within.

1-4) The 4th memory buffer space is a memory for simply outputting the final result calculated by the digital IC chip to an external receiver. The primary memory buffer space is used to synchronize the external audio signal with the digital IC chip system clock to provide sampling time. At the same time, the audio signal finally processed from the 4th memory buffer space is synchronized with the digital IC chip system clock and output at the sampling time. Due to this invention, the above-mentioned disconnection phenomenon is fundamentally solved.

2. This time, as an example, the performance of a digital hearing aid in a digital IC chip with 128 input voice signal data (1, 2, ... 128) contained in 128 secondary memory buffer spaces for 2 msec. I will explain the process of executing all necessary operations in parallel.

2-1) Move 128 data in the secondary buffer memory space to the fifth memory buffer space (pre-time buffer). When fast fast pre-conversion is performed from 128 time domain real data, x (n), in the 5th buffer memory space, 128 frequency domain complex data, X (n), is calculated. N = 1 to 128 here. The calculated 128 frequency-domain complex data are stored in a sixth order memory buffer space (free frequency buffer). In the calculated 128 frequency domain complex data, only 65 from the first 1 to 65 are stored in the 7th memory buffer space (linear buffer 1) from the 6th memory buffer space. Since the seventh memory buffer space needs to store complex numbers, it consists of a total of 130 memory buffer spaces in two groups of 65 each, and is divided into 65 real numbers and 65 imaginary numbers.

이고 위상은

이다.2-2) The digital hearing aid is deteriorated by performing the most appropriate fitting within the hearing threshold threshold dynamic range based on the hearing test result of the hearing impaired person who measured the hearing threshold in logarithmic units (dB HL) as a function of the frequency of a certain unit. The primary purpose is to correct the established hearing threshold. Therefore, it is very efficient to process all the operations necessary to realize the performance of the digital hearing aid in dB after the fast IC conversion from the digital IC chip to the final fast reverse conversion. To do this, amplitude and phase are first calculated from complex data consisting of real and imaginary numbers in the 7th memory buffer space, and converted into amplitude and phase instead of real and imaginary numbers. If real and imaginary are x and y respectively, the amplitude is

And phase is

to be.

2-3) Convert 65 amplitude data of 7th memory buffer space in dB and restore it.

20 x log ₁₀ (x)

2-4) The 65 amplitude data in the 7th memory buffer space in dB is the sound pressure signal of the voice detected by the microphone of the hearing aid and is generally calculated to be 0 dB or less. Absolute sound pressure correction should be applied to this value to convert the actual sound pressure into units of dB SPL (Sound Pressure Level) .For absolute sound pressure correction, the absolute sound pressure correction value (in dB SPL units) of the frequency function obtained from the reception sensitivity measurement of the microphone is 7th order. The frequency is added to 65 amplitude data (in dB) in the memory buffer space.

2-5) Convert 65 amplitude data (in dB SPL unit) of the 7th order memory buffer space in 2-4) above to dB HL unit. To do this, the loudness curve of the frequency function is used.

2-6) After the spectral modulation algorithm is performed using the 65 amplitude data (in dB HL) of the 7th order memory buffer space of 2-5) above, the final result is transferred to the 8th order memory butter space (log buffer). Store as 65 amplitude data (in dB HL).

2-7) Convert 65 amplitude data (in dB HL) from the 8th order memory buffer space to dB SPL. To do this, we use a loudness inverse curve of a frequency function that is commonly used.

2-8) Convert 65 amplitude data (in dB SPL units) from the 8th order memory buffer space to dB units. To do this, subtract the absolute sound pressure correction value (in dB SPL units) of the frequency function from the 65 amplitude data (in dB SPL units) in the 8th order memory buffer space for each frequency.

2-9) 65 amplitude data of 8th order memory buffer space stored in dB unit is converted into linear unit and restored.

10 ^{0.05 * x}

2-10) 9th-order memory by converting from 65 amplitude data (linear unit) in the 8th memory buffer space and 65 phase data (linear unit) stored in the 7th memory buffer space into a complex unit by a conventional calculation method. Move and store in buffer space (linear buffer 2).

x + j * y = Amp x cos (phase) + j * Amp x sin (phase)

2-11) The 65 complex data of the 9th memory buffer space is extended to 128 complex data according to the above-mentioned expectation method, and then moved and stored in the 6th memory buffer space.

2-12) It converts the complex data from 128 frequency domains of the 6th memory buffer space into 128 time domain real data by fast free inverse transform and transfers them to the 5th memory buffer space.

2-13) 128 time domain real data (final digital signal processed data) of the 5th memory buffer space is moved and stored in the 3rd memory buffer space.

3. Next, for example, the process of executing the spectrum modulation algorithm operations in 2-6) in parallel will be described. The spectral modulation algorithm is divided into two parallel processing.

One is a spectral amplitude modulation algorithm that changes the shape of the amplitude spectrum curve of the speech, such as spectral compression, spectral squelch, and spectral equalization. The other is to arbitrarily titrate the shape of the specific amplitude spectra resulting from narrowband noise or acoustic feedback. It is an automatic spectral noise cancellation algorithm. The frequency spectrum is 0Hz, 125Hz, 250Hz, 375Hz, 500Hz, 625Hz, 750Hz, 875Hz, when the sampling frequency is 16000Hz for 65 amplitude data (dB HL units) stored in the 7th memory buffer of an embodiment of the present invention. 1000 Hz, 1125 Hz, 1250 Hz, 1375 Hz, 1500 Hz, 1625 Hz, 1750 Hz, 1875 Hz, 2000 Hz, 2125 Hz, 2250 Hz, 2375 Hz, 2500 Hz, 2625 Hz, 2750 Hs, 2875 Hz, 3000 Hz, 3125 Hz, 3250 Hz, 3375 Hz, 3500 Hz, 3625 Hz, 3750 Hz, 3875 Hz, 4000 Hz, 4000 Hz, 4000 Hz 4125 Hz, 4250 Hz, 4375 Hz, 4500 Hz, 4625 Hz, 4750 Hz, 4875 Hz, 5000 Hz, 5125 Hz, 5250 Hz, 5375 Hz, 5500 Hz, 5625 Hz, 5750 Hz, 5875 Hz, 6000 Hz, 6125 Hz, 6250 Hz, 6375 Hz, 6500 Hz, 6625 Hz, 6750 Hz, 6875 Hz, 6000 Hz, 7125 Hz, 7125 Hz Frequency intervals are determined by 7250 Hz, 7375 Hz, 7500 Hz, 7625 Hz, 7750 Hz, 7875 Hz, and 8000 Hz.

To describe a spectral amplitude modulation algorithm that changes the shape of an amplitude spectrum curve such as spectral compression, spectral squelch, and spectral equalization, first, the threshold level curves by hearing test must be known. The threshold level curve is divided into three. The threshold level curve is measured and prepared according to the auditory response of the subject while the sine wave signal of 64 pure tone frequencies except 0 Hz is heard close to the ear drum through the pressure-compensated earphone. Hearing threshold level curves represent the degree of hearing that is almost audible to the pure tone of an individual frequency. This means that you will not hear a sound pressure that is lower than the hearing threshold level curve. The discomfort threshold level curve is the degree of auditory response to sound at a sound pressure that is uncomfortable for the subject to hear. Sound pressure at higher intensity than the offended threshold level curve means that the sound pressure is high enough to be unpleasant to hear. The stable threshold level curve indicates the degree of auditory response at which the sound is most comfortable to the subject for individual frequency pure tones. The intensity of the sound felt most comfortable by the subject varies depending on the frequency. The subject selects the threshold level most comfortable according to the hearing preference felt by the subject. The stable threshold level curve therefore naturally determines the spectral equalization desired by the subject. In addition, the subject may vary the loudness of undesired frequencies, such as bass or treble, and change the self-stable threshold level curve according to the subject's choice. Therefore, the external sound is small to the deaf and hard of hearing, and when the external sound is amplified by the stable threshold level curve of each deaf person, it is most comfortable. Since all three threshold level curves are measured in dB HL, the 65 amplitude data stored in the 7th memory buffer are stored in dB HL. The difference between the unpleasant threshold level curve and the hearing threshold level curve is called the hearing dynamic range, which varies with frequency as a function of frequency, and the hearing dynamic range is the lowest and highest physical limit at which hearing loss can hear external sounds. Set it to a range.

The 65 amplitude data (in dB HL units) stored in the 7th memory buffer are voice amplitude spectra representing sound pressure strength of the external voice signal. The external speech amplitude spectrum forms a different pattern of curves every time between the hearing threshold level curve, the stable threshold level curve, and the discomfort threshold level curve of the hearing impaired at each frequency.

We also considered sound with a small intensity below the hearing threshold level curve. Everyday conversation sounds make up syllables (eg 'chapters') with phonemes (eg 'ㅈ', 'ㅏ', 'ㅇ'), and syllables add words to make up words (eg 'roses') do. And syllables can be divided into primary, neutral, and final. However, in general, vowels have higher acoustic energy and frequencies closer to bass than consonants, while consonants have lower acoustic energy and higher frequencies than vowels. In particular, most deaf people miss the consonant because the acoustic energy of the consonant is small. Therefore, a digital hearing aid must be able to detect at least the consonant of the dialogue and amplify the speech amplitude of the frequency by the hearing threshold level curve so that the consonant can hear the consonant. However, if you amplify even the small noises of the surroundings, not even the conversational sounds, it is difficult to discriminate the speech of the conversational sounds to be heard by the surrounding noises. Therefore, it is necessary to distinguish the difference between the small noise and the dialogue sound. In the present invention, the spectrum of the voice signal continuously input for 200 msec corresponds to the sound threshold of the acoustic energy of the initial consonant of the normal conversation sound ( If it is smaller than the consonant threshold level curve (in dB HL), it is regarded as a small noise around it and we want to perform attenuation rather than amplification. This is called automatic spectral squelch adjustment. In other words, if the ambient noise sound signal of a smaller intensity than the consonant initial of the conversation sound is input through the microphone of the digital hearing aid, the time of 200 msec is observed. If the sound of small intensity is still input even during 200 msec, it is regarded as the ambient noise. Instead of amplifying the hearing aids, attenuation is performed. This requires a timer to observe the spectral squelch at 200msec intervals. The reason for having an observation time of 200 msec is that a syllable time of a typical conversation sound is a maximum of 200 msec. The digital IC chip of a digital hearing aid normally amplifies the input voice with an intensity smaller than the hearing threshold level curve, but when input continues to be smaller than 200 msec, attenuation is continuously amplified by performing attenuation instead of amplification immediately. In order to avoid exposure to ambient noise, a hearing impaired person wearing a digital hearing aid may focus on conversational sounds.

If it is judged to be normal conversation sound, therefore, the external voice amplitude spectrum is automatically adjusted for each frequency so that the center value approximates the stable threshold level curve within the hearing dynamic range between the hearing threshold level and the discomfort threshold level curve of the deaf person. If you adjust it, the deaf person can hear the sound most comfortably in their hearing. To do this, one must amplify the speech amplitude spectrum randomly at another frequency or attenuate it at another frequency.

The difference in energy levels between the initial consonant and the neutral vowels of a normal conversation sound is 45-50 dB for a typical conversation sound. However, the range of hearing dynamics, which is the difference between the hearing threshold level and the discomfort threshold level of the hearing impaired, varies with frequency. As the degree of hearing loss increases from mild to moderate, medium, high, and deep hearing loss, the hearing loss becomes narrower. For example, the hearing dynamic range of highly hearing impaired persons may be reduced to 30 dB. Therefore, it is necessary to change the amplitude spectrum of speech according to frequency so that the dynamic range of dialogue sound falls within the hearing dynamic range of the hearing impaired person. This is called spectral compression.

비율이 나오므로, 예를 들면, 주파수에서 입력 음성 스펙트럼의 진폭이 x=50dB HL이라면,

공식에 의해, 출력되어야 할 진폭은 y=76.6667dB 이며, 원래 입력 진폭이 50dB HL이므로, 증폭 정도는 y-x=16,6667dB로 결정된다.Spectral compression sets the dynamic range of the dialogue sound in advance, and then writes an algorithm to vary the degree of amplification at a constant rate differently according to the measured hearing dynamic range of the hearing impaired person. In one embodiment, the dynamic range of the dialogue sound is set at 25 dB HL to 70 dB HL at a certain frequency, and the hearing dynamic range of the hearing impaired person at the same frequency is 60 dB HL to 90 dB HL. then

Ratio, so, for example, if the amplitude of the input speech spectrum at frequency is x = 50 dB HL,

By the formula, the amplitude to be output is y = 76.6667 dB, and since the original input amplitude is 50 dB HL, the amplification degree is determined to be y - x = 16,6667 dB.

When analyzing the frequency characteristics of a typical dialogue tone with a fast free-switch, the vowel threshold level among the dialogue sounds is as high as 75dB HL at low and 70dB HL at high, and the consonant threshold level is as high as 35dB HL at low and 25dB at high. Lowered to HL. This indicates that the dynamic range of dialogue is about 40 dB at low and about 45 dB at high. Hearing threshold level is lower than the consonant threshold level of the conversational sound so that the normal person understands all the conversation sounds, but the hearing impaired person raises the consonant threshold level of the conversational sound at least the hearing threshold level because the hearing threshold level is higher than the consonant threshold level of the conversational sound. Amplification of digital hearing aids is inevitable. On the other hand, the vowel threshold level of dialogue sounds may be higher than the hearing threshold level for light hearing impaired persons and lower than the hearing threshold level for moderate hearing impaired persons. Therefore, the input voice must be amplified or attenuated differently for each frequency so that the dynamic range of the dialogue sound falls within the hearing dynamic range of the hearing impaired person.

The present invention proposes a new algorithm for matching the range of dialogue dynamics between the consonant threshold level and the vowel threshold level of a conventional dialogue tone to the hearing dynamic range between the hearing threshold level and the discomfort threshold level of a hearing impaired person. This is divided into dynamic range compression algorithm among spectral amplitude modulation algorithms. The proposal of the present invention proposes to amplify the consonant threshold level of a typical conversation sound by 5 dB (= A) more than the hearing threshold level. This makes it possible for the deaf person to hear the consonant consonant of the dialogue sound most easily missed. It is proposed to amplify the vowel threshold level of the normal conversation sound by 10 dB (= B) more than the stable (desired) threshold level. This amplifies the vowels only to the extent that the hearing impaired does not feel uncomfortable, making it possible to comfortably wear digital hearing aids. While this suggestion is not difficult for people with mild to moderate hearing loss, it can be very narrow, below 10 dB (= B), between the stable (desired) threshold level and the unpleasant threshold level for moderate to high hearing loss individuals, in which case the typical dialog It is proposed to amplify the vowel threshold level by only the offending threshold level. This change in amplification applies differently for individual frequencies.

4. A spectral amplitude modulation algorithm that changes the shape of an amplitude spectrum curve, such as spectral compression, spectral squelch, and spectral equalization, mentioned above, is a frequency range of 64 speech amplitudes in the seventh-order memory buffer space. This is a programming method that approximates the hearing dynamic range between the hearing threshold level curve and the displeasure threshold level curve measured by the deaf person. The dynamic range compression algorithm method to achieve this is as follows.

4-1) Storing the discomfort threshold level curve measured by the hearing test from the hearing impaired in the 10th memory buffer space (discomfort threshold buffer) having 65 addresses from 2 to 65 times corresponding to the frequency of 125 Hz to 8000 Hz. (1 stores 0 values), and similarly stores the stable (desired) threshold level curve measured by the hearing test from 2 to 65 times in the 11th memory buffer space (desired threshold buffer) with 65 addresses ( 1 stores the measured hearing threshold level curve in the 12th memory buffer space (hearing threshold buffer) with 65 addresses (number 1 stores 0 values).

4-2) Stores the vowel threshold level curve of a typical conversation sound in the 13th memory buffer space (collection threshold buffer) with 65 addresses from 2 to 65 corresponding to frequencies from 125 Hz to 8000 Hz (1 is Similarly, the consonant threshold level curve of the conversation sound is stored in the 14th memory buffer space (consonant threshold buffer) having 65 addresses from 2 to 65 (1 stores the 0 value). In addition, binary 0 values from No. 2 to 65, corresponding to frequencies from 125 Hz to 8000 Hz, are stored in the 15th memory buffer space (squeeze buffer) with 65 addresses (1 is 0).

4-3) If the audio signal amplitude spectrum data, which has been moved in the 7th memory buffer space at 2 msec intervals, is immediately moved to the 8th memory buffer space for each frequency, amplitude spectrum modulation is not performed at all. For spectral amplitude modulation, we first copy 65 binary data from the 7th memory buffer space into the 8th memory buffer space (log buffer) with 65 binary addresses. If the speech amplitude spectrum shown in the eighth memory buffer space is greater than the discomfort threshold level stored in the tenth memory buffer space, it is immediately corrected to the discomfort threshold level. This is to automatically limit the maximum output unpleasant for hearing impaired.

4-4) Since the 200msec time interval must be observed continuously for the above-mentioned spectrum squelch function, a timer built in the digital IC chip or operated by programming is set automatically at the time of digital IC chip initialization. . 65 binary data of the 7th memory buffer space is stored as an average value in the 15th memory buffer space. In other words, the pre-stored frequency-specific data (numbers 1 to 65) in the 15th memory buffer space are summed with new data (numbers 1 to 65) newly input from the 7th memory buffer space, and then divided by 2 again. The average value is again stored in the 15th memory buffer space for each frequency from 1 to 65 times. This allows the 15th order memory buffer space to be used to observe the average amplitude spectral level of an external speech signal. When the timer delivers an interrupt signal every observation time (200 msec) by the interrupt method, if the amplitude spectral strength of the 15th memory buffer space is compared and summed by frequency with the 14th memory buffer space for each frequency, the situation is smaller than the appropriate threshold. Determining that no audio signal is input from outside (squeeze condition occurs), set the contents of the 8th memory buffer space to 1dB HL value, reinitialize the 15th memory buffer to binary 0 value, and then Terminate the spectral amplitude modulation (dynamic range compression) algorithm.

4-5) If the squelch does not occur, continue to the next step.

As described above, the range of the dialogue sound dynamics between the consonant threshold level (14th memory buffer) and the vowel threshold level (13th memory buffer) of a typical conversation sound is determined by the hearing threshold level (12th memory buffer) and the discomfort threshold of the hearing impaired person. As a dynamic range compression algorithm that tries to fit the hearing dynamic range between levels (10th memory buffer), in the present invention, the consonant threshold level (14th memory buffer) of a normal conversation sound is 5dB above the hearing threshold level (12th memory buffer). It is proposed to amplify by (= A), and the vowel threshold level (13th memory buffer) of the normal talk sound is 10dB (= B) more than the stable (desired) threshold level (11th memory buffer) or the discomfort threshold level. It was proposed to amplify only up to (10th order memory buffer). This is calculated separately for the frequencies of 2 to 65 memory addresses.

In one embodiment, the consonant threshold level of the conversation sound is a, the vowel threshold level of the conversation sound is b, the displeasure threshold level is c, the stable (desired) threshold level is d, and the hearing threshold level is e at one frequency. Let's say. The amplification factor is determined by y as in the following equation.

Where f is the value of d + B or c, prefering values less than or equal to c. As an example, A is 5dB and B is 10dB, but since the customized digital hearing aid is customized, the A and B values can be adjusted according to the hearing condition of the customer. x is the speech spectral amplitude of the seventh order memory buffer space and y is the amplified speech spectral amplitude to be stored in the eighth order memory buffer space. In this way, a dynamic range compression algorithm is performed for each frequency.

This paper describes a spectral noise cancellation algorithm that automatically and arbitrarily adjusts the shape of the singular amplitude spectra caused by the narrowband noise or acoustic feedback. The voice signal amplitude data (No. 1 to No. 65) newly input and stored in the 7th memory buffer space are the instantaneous voice signal spectral amplitudes input at intervals of 2 msec. The amplitude variation of the instantaneous speech signal spectrum is large, and when averaged over time, the spectrum flattens and becomes stable over time.

Then, when acoustic feedback (howling) or narrowband noise occurs, the amplitude of the frequency band corresponding to a specific howling or narrowband noise suddenly increases from the mean value continuously calculated over time. Such drastic changes are likely to occur mainly in the high frequency band. Such acoustic feedback or narrowband noise is awkward to the normal as well as the hearing impaired.

Therefore, if it is observed that such a sudden change in the amplitude spectrum occurs in the narrow band, it should be regarded as a kind of noise and the amplitude of the corresponding frequency band should be lowered. The use of digital hearing aids such as automatic feedback cancellation or narrowband noise canceling makes it easier for wearers to use their hearing aids.

5. Now we describe the process of implementing the just mentioned spectral noise cancellation algorithm function.

5-1) Stores the audio signal amplitude spectrum data (numbers 1 to 65) that are newly input and stored continuously in the 7th memory buffer space at 2 msec intervals in the 15th memory buffer space (squench buffer) at 65 addresses. At the same time, the 16th memory buffer space (feedback buffer) of 65 addresses is averaged and stored in parallel processing. The 16th order memory buffer space was initialized to zero values at the same time as the 15th order memory space. The amplitude data for each frequency (numbers 1 to 65) in the 16th memory buffer space are summed with the new amplitude data (numbers 1 to 65) newly input from the seventh memory buffer space, and then divided by two. The average value is again stored in the 16th memory buffer space for each frequency from 1 to 65 times. This allows the 16th order memory buffer space to be used to observe changes in the average amplitude spectral level of the external speech signal. Unlike squelch control, feedback control does not require a timer. This is because as soon as acoustic feedback or narrowband noise occurs, the amplitude of the frequency band must be lowered. Therefore, if any amplitude of the spectral amplitude stored in the 16th order memory buffer is larger than the initial threshold, that is, when acoustic feedback or narrowband noise occurs, the amplitude spectral value at that frequency is immediately stabilized (desired) threshold. Lower to level (11th memory buffer).

In this way, a spectral noise cancellation algorithm modulates the amplitude spectrum of a signal with an unusual spectrum that differs from the speech of undesired abruptly occurring ambient noise such as acoustic feedback or narrowband noise.

In the drawings, reference numeral 30 denotes a battery door.

In the above description, the specific embodiments have been shown and described, but the present invention is not limited thereto, for example, by those skilled in the art without departing from the concept of the present invention. Of course, various design changes are included in the claims of the present invention.

10: ear shell 30: battery door

Claims (3)

During the 0.0625msec sampling time as soon as 32 new voice signals are input through the input buffer, it processes digital signals for 128 input signals per frame but does not cause some signal dropouts due to signal processing time by the CPU. Performing a programming algorithm for moving the 32 final voice signals, which were input first, out of the last 128 frame data for which all spectrum modulation algorithm signals have been processed to the output buffer, and then outputting the 32 output buffers in synchronization with a receiver at 0.0625 msec sampling time. Wow;
After performing the programming algorithm, after converting only 65 complex data into log unit (dB HL) after fast free-switching, the spectrum amplitude modulation signal processing is performed, and then from log unit (dB HL) to 128 complex units of expected values Performing a fast free inverse transform algorithm which reconverts and computes 128 completed final output speech data with fast free inverse transform;
If the spectrum of the speech signal continuously input for 200 msec after the fast free inverse transform algorithm is smaller than the consonant threshold level curve of the conversation sound, it is regarded as ambient noise and the attenuation is not exposed to the amplified ambient noise by attenuation rather than amplification. And performing a spectral amplitude modulation algorithm for concentrating only on speech-oriented voices. The method of claim 1, wherein the frequency of amplifying the input speech signal in the step of performing the spectral amplitude modulation algorithm is varied for each frequency, and the consonant threshold level of the normal conversation sound is amplified by 5 dB more than the hearing threshold level of the hearing impaired person. By amplifying the vowel threshold level of the conversational sound by 10 dB more than the stable (desired) threshold level of the hearing impaired person, or by only the maximum unpleasant threshold level, the dynamic range of the normal speech sound is adjusted to the hearing dynamic range of the hearing impaired person, Signal processing method of a digital hearing aid, characterized in that to improve the discrimination ability of speech in accordance with the wearing of the hearing aid. The method of claim 1, wherein in the performing of the spectral amplitude modulation algorithm, an average value of the speech amplitude spectrum is calculated and stored, and the usual spectral shape is monitored. In the event of extremely loud acoustic feedback or narrow-band noise, the user may arbitrarily lower the amplitude of the corresponding frequency band so that unpleasant acoustic feedback (howling) or unusual external noise is not exposed to the hearing impaired person wearing the digital hearing aid. Signal processing method of a digital hearing aid.

Images