KR100372576B1 - Method of Processing Audio Signal - Google Patents

Abstract

Depending on the value set by the user, the attenuation is differentially amplified by applying a higher gain ratio relative to the strong sound. An audio signal processing method is disclosed. For differential amplification, we substitute sample data into the monotonic increment function and replace it with the value of that function. If necessary, select a range to apply differential amplification and differentially amplify only the samples included in the range. For time-base conversion, the original signal is read in units of frames and superimposed on the time-base conversion signal, and the overlapped part is the basic difference value between the start time adjustment rate α specified by the user and the starting point of each frame of the original signal x (·). It depends on Sa. In addition, the overlapping parts are summed by applying weights. The original signal to read a predetermined check, based on the primary differential value Sa of the inter-frame entry point when all range (window) optimal, based on the similarity of the waveform between the original signal x (·) and the time-base conversion signal y (·) in the k _m Correlation is detected and its position is determined as the synchronization delay Km. The differential amplification can be performed before or after the time base transformation on the frame data read for the time base transformation.

Description

Method of Processing Audio Signal

The present invention relates to an audio signal processing method and apparatus therefor. More specifically, the present invention relates to a method for processing an audio signal in which the reproduction time and intensity of the audio signal are simultaneously changed in real time according to a value set by a user.

Sometimes it is difficult to hear the playback sound output through the audio playback device such as voice or song. There are many possible causes for this. The reason may be that the original tone is weakly recorded or recorded at a very high speed, but it is also possible that the listener's hearing is weak and he / she does not hear it well, or that the hearing is fundamentally low due to the foreign language being played. This can be

In particular, in learning foreign language listening ability, one solution may be to reinforce the weak sound strength, which is difficult to hear as a method of enhancing the listening ability. In general, syllabic language such as English has a linguistic characteristic that can convey more information by accent and accent, unlike Korean or Japanese. Omitting or attenuating some of the various pronunciations in a word does not have much problem in conveying its meaning, so syllable languages, especially English pronunciations, are often treated as mute or weak. In contrast, in the case of nonsyllabic languages such as Korean or Japanese, communication through the voice is mainly responsible for the meaning included in each word, so the accent or accent does not take much weight. The pronunciation feature of non-syllable languages is that each word is pronounced correctly, and it is rarely pronounced as a silent or weak sound. It is very difficult for people who use non-syllable languages such as Korean or Japanese to listen to the accented syllable parts of English pronunciations, such as silent or weak sounds. As a result, those who are familiar with Korean or Japanese pronunciation systems needed to repeat listening lessons in order to understand English pronunciation correctly.

It is very difficult to artificially add silence, so that it is possible to enhance the strength of the weak sound when the audio signal is reproduced. However, known gain control methods of audio signals are methods that apply the same amplification rate uniformly without depending on the intensity of sound, but rather are not actually employed because they degrade the quality of the reproduced sound. In other words, according to these methods, if the amplification of the weak sound level is applied to the desired level by applying the same gain rate uniformly without distinguishing the weak sound or strong sound, even the strong sound is strengthened and the strong sound part is heard as a loud and torn sound. The quality of the reproduction sound is rather worse.

On the other hand, if the original normal playback speed feels relatively fast, it may be necessary to make the playback speed slower or vice versa. Various techniques are known in the art for varying the playback speed. For example, since the overlap-addition (OLA) was introduced by Roucus and Wilgus in 1985, it changed to the synchronous overlap addition method (SOLA), the OLA method based on waveform similarity, that is, the WSOLA method. It is developing. In order to change the playback speed, the SOLA algorithm overlaps the audio data at regular intervals by using a window of a constant size, cuts out the corresponding block, and adds the audio signal by rearranging the blocks on the time axis in response to the required speed change. Processing. In order to prevent sound degradation, a technique of synthesizing two block signals by shifting the similarity of waveforms of overlapping portions by an interval corresponding to the largest value may be applied.

However, the technology of changing the reproduction speed of the audio signal can maintain good sound quality even after processing, and furthermore, it is highly desirable if it can be simultaneously processed in combination with a technique of varying the sound intensity.

It is a first object of the present invention to provide an audio signal processing method that amplifies a weak sound relatively stronger than a strong sound based on a value determined by a user, thereby processing the weak sound more easily. .

It is a second object of the present invention to provide an audio signal processing method which can vary the reproduction time according to a value determined by a user of an original audio signal, and at the same time, further enhance the strength of the weak sound compared to the strong sound.

1 is a block diagram showing the configuration of an apparatus for realizing an audio signal processing method according to the present invention.

2 is a flowchart schematically illustrating an audio signal processing method according to the present invention.

Fig. 3 is a flowchart showing an algorithm for performing time-base conversion processing on the original audio signal according to the reproduction time adjustment rate? Set by the user.

FIG. 4 is a reference diagram for explaining a principle of obtaining the time-axis conversion signal y (·) by increasing or decreasing the length of the original audio signal x (·) corresponding to the size of the reproduction time adjustment rate α set by the user. .

5 is a flowchart for explaining an algorithm for differentially amplifying by applying amplification higher than weak sound.

6 is a graph illustrating various examples of the differential amplification function.

7 shows waveform diagrams of the original signal, the time-base converted signal, and the differentially amplified signal, respectively.

** Explanation of symbols for main parts of drawings **

10: memory

12: playback time adjustment unit

14: differential amplifier

16: filter section

18: buffer part

20: user interface

22: audio output unit

24: speaker

In order to achieve the first object, the present invention provides a method for processing a digital audio signal expressing the sound intensity over time as an integer stream,

Reading the differential amplification factor β specified by the user from the user interface; And

Each sample data of the integer stream is amplified by applying a variable value of a predetermined differential amplification function f (x _i ) corresponding to the differential amplification ratio β, and in particular, the differential amplification function f (x _i ) is weak to strong. It provides a method for processing an audio signal, characterized in that it comprises a differential amplification step that is a function of applying a relatively low amplification gain gradually differentially compared to strong sound.

Preferably, the differential amplification function f (x _i ) is selected from the monotonically increasing function.

The audio signal processing method preferably further includes performing low pass filtering on the packet data when the differentially amplified sample data is collected by the packet data amount.

In one selectable method, the differential amplifying step includes: comparing the sample data x _i with a threshold value x _th ; And as a result of the comparison, when the respective sample data x _i is greater than the threshold value x _th , the differential amplification function f (x _i ) is f (x _i ) = x _i ^{1 / β} , and each sample data x _i is If the threshold value is smaller than x _th , the differential amplification function f (x _i ) becomes f (x _i ) = x _i ^β and includes performing the differential amplification process.

As another selectable method, the step of comparing the sample data x _i of the differential amplification step with a threshold value x _th and its magnitude; And as a result of the comparison, the differential amplification function f (x _i ) is performed as f (x _i ) = x _i ^{1 / β} only when each sample data x _i is larger than the threshold value x _th . Performing the steps.

On the other hand, in order to achieve the second object, the present invention is to adjust the reproduction time of the digital audio signal according to the user's designated value by using the digital audio signal expressing the sound intensity with respect to time as an integer stream as an input signal The present invention provides an audio signal processing method characterized in that the gain control by the differential amplification and processing in parallel.

The audio signal processing method reads the reproduction time adjustment rate α and the differential amplification ratio β, which are input values specified by the user, from the user interface, wherein the reproduction time adjustment rate α is longer than the normal reproduction time. As a short adjustment, it is determined within the range of α _min ≤ α ≤ α _max , and the differential amplification ratio β is for reinforcing the weakness of the audio signal relatively more than the strong sound, and is determined within the range of β _min ≤ β ≤ β _max . Losing set value reading step;

The original audio signal x (·) stored in the first memory is continuously read in units of N samples, and the next frame F _m is smaller than the N at the start of the previous frame F _m-1 . A data reading step of reading a predetermined number of samples in duplicate between adjacent frames by reading from a sample of a position where a sample of the skipped position is skipped;

The next frame F _{m is added} to the second memory so as to overlap the previous frame F _m-1 already recorded as the time-base conversion signal y (·) by a predetermined number of samples, and the samples of the overlapped portion are weighted, added, and not duplicated. The portion is added as it is, and the number of samples of the overlapped portion is added in such a manner as to increase or decrease in correspondence with the size of the reproduction time adjustment rate α so that the time axis for obtaining a signal y (·) whose time axis length is converted in proportion to the reproduction time adjustment rate α is obtained. Conversion step; And

Sample data x _i of each frame is amplified by substituting a predetermined differential amplification function f (x _i ) corresponding to the differential amplification ratio β, in particular, and the differential amplification function f (x _i ) is amplified from weak to strong. The gain is a relatively low applied function, and includes a differential amplification step that differentially intensifies the weakness compared to strong sound.

If necessary, if the sample data of each frame read in the data reading step is data that is not normalized, further includes a normalization step of normalizing it.

In the audio signal processing method, preferably, a plurality of output buffers are provided, and a plurality of frames of a signal processed through the time-base conversion step and the differential amplification step are bundled into one packet, and among the plurality of output buffers. In particular, the output buffer for writing the packet data is replaced in a circular manner, and the packet data recorded in the plurality of output buffers in parallel is the packet data of the output buffer prior to the recording order including the speaker first. It further includes a data output step of controlling to be output to the audio output unit.

The audio signal processing method may further include a filtering step of removing noise by performing low pass filtering on the differentially amplified sample data for each predetermined data amount.

Hereinafter, an audio signal processing method according to the present invention will be described in more detail with reference to the accompanying drawings.

1 is a block diagram showing the configuration of an apparatus for realizing an audio signal processing method according to the present invention. The audio signal processing apparatus of FIG. 1 includes a memory 8 for receiving digital audio data and storing the same, and a playback time adjusting unit for processing a signal processing to read audio data from the memory 8 and vary the playback time ( 12) and a differential amplifier 14 for amplifying a weak sound with a relatively larger gain than the strong sound. The reproduction time adjustment section 12 increases or decreases the reproduction time of the audio signal based on the reproduction time adjustment rate α. The differential amplifier 14 performs the differential amplification process in accordance with the differential amplification ratio β. The reproduction time adjustment rate α and the differential amplification rate β are provided from the user interface unit 20, respectively. That is, the user interface unit 20 monitors the user's operation and converts the user's setting value for the reproduction time of the audio signal and the reinforcement of the attenuation into a reproduction time adjustment rate α and a differential amplification rate β, thereby adjusting the reproduction time adjustment unit. (12) and the differential amplifier (14). In some cases, the user interface unit 20 further provides a threshold value x _th of the sound intensity to which the weakening process is applied.

The digital audio data provided to the memory 10 expresses the sound intensity with respect to time as an integer stream, and is preferably data normalized by the maximum value. In the case of non-normalized data, the playback time adjusting section 12 or The differential amplifier 14 normalizes in advance before signal processing. The source of digital audio data may be a hard disk or a CD-ROM which stores an audio file (eg, an audio file in wav or avi format), or may be an online input means such as a microphone. In order to process the output signal of the microphone, since the output signal is an analog signal, the process of digitizing it must be preceded.

The audio signal processing device also receives the data after the reproduction time and the differential amplification processing, and performs a low pass filtering to remove the noise in the high frequency band, and buffers the data processing speed difference between the front and rear ends to output the data to the output stage. An output buffer unit 18 for smooth supply is further included, and an audio output unit 22 for converting a digital audio signal into an analog audio signal and outputting sound through the speaker 24.

The reproduction time adjusting unit 12 and the differential amplifying unit 14 constitute the signal processing unit 10, which may be implemented in software on a digital signal processing (DSP) chip (not shown) together with the filter unit 16. Computer devices are also implemented in software. The audio output unit 22 includes an analog / digital converter, a signal amplifier, and the like, and in the computer device, the sound card may correspond to this.

2 is a flowchart schematically illustrating an audio signal processing method according to the present invention.

The user can set the desired reproduction time adjustment rate α and the differential amplification rate β through predetermined operation means. The reproduction time adjustment rate α is for adjusting the reproduction time of the audio signal longer or shorter than the normal reproduction time, and is determined within the range of α _min ≦ α ≦ α _max . If the value of α is 1, it is reproduced at the same speed as the original normal reproduction speed. If the value is less than 1, the reproduction time is shortened (ie, the reproduction speed is increased) because the amount of the digital signal is reduced. If the value is greater than 1, the playback time increases and the playback sound is slow. In the case of the audio signal, it can be confirmed through actual implementation that α can be reproduced with sufficiently good sound quality even if it is set within the range of 0.7 or more and 3 or less. The differential amplification ratio β is for reinforcing the weak sound of the audio signal relatively more than the strong sound, and is determined within the range of β _min ≦ β ≦ β _max . In the case of the audio signal, β also has a value within the range of 1 to 3, it can be confirmed through the actual implementation that it can be reproduced with a sufficiently good sound quality.

When the user inputs the reproduction time adjustment rate α and the differential amplification ratio β by using a predetermined input means such as a keyboard, a mouse, or another type of input means, the user interface 20 monitors the input value to monitor the input value. Provide each to the differential amplifier 14 (step S10). When these values are provided during the processing of the audio signal, the reproduction time adjustment section 12 and the differential amplifier section 14 are interrupted, and the existing values are replaced by the newly provided reproduction time adjustment rate α and the differential amplification ratio β.

The memory 8 stores the original digital audio signal x (·) before processing, which is read by the signal processing section 10 (step S12). The unit of reading signal is frame. The frame includes N sample data. In order to adjust the playback time, the next frame is not read from the end of the previous frame, but adjacent frames are read in duplicate. That is, the next frame F _m reads out a predetermined number of samples repeatedly between adjacent frames by reading from the sample of the position where the number of samples less than N is skipped from the start position of the previous frame F _m-1 . A more detailed description thereof will be described later.

For each frame read through the data reading step (step S12), the signal processing unit 10 performs a differential amplification step (step S14) and a time axis conversion step (step S16), that is, a playback time conversion step. The differential amplification step (S14) and the time-base conversion step (S16 step) may process the audio signal in the order of processing the differential amplification first and then processing the time-base conversion as shown in the flowchart of FIG. 2, but the reverse order is also possible.

The differential amplification step (step S14) is a step of amplifying by substituting the sample data x _i of each frame into a predetermined differential amplification function f (x _i ) corresponding to the differential amplification factor α. Here, the differential amplification function f (x _i ) is a function in which the amplification gain is relatively low, especially from weak to strong. Therefore, by converting the audio data using this function, it is possible to differentially strengthen the weak sound compared to the strong sound.

In the time-base conversion step (step S16), the next frame F _m is converted into a time-base conversion signal y (· which may be physically a specific area of the memory 8 or an internal memory of the signal processing unit 10). ) Is recorded so as to overlap with the previous frame F _m-1 already recorded as a predetermined number of samples. Here, the sample of the duplicated part is weighted and added, and the non-duplicate part is added as it is. The number of samples of the overlapping portion is added in a manner of increasing and decreasing corresponding to the size of the reproduction time adjustment rate?. Thus, the original signal x (·) has its time axis length converted in proportion to the reproduction time adjustment rate α.

In this way, the audio data converted into the reproduction time adjustment rate and the differential amplification ratio set by the user by the differential amplification and time-axis conversion processing are filtered by the filter unit 16 and recorded in the buffer unit 18 (step S18). . Undesired noise can be introduced during the differential amplification and time-base conversion process. Since the noise usually has a high frequency component, the low pass filtering process removes the noise. Filtering may be performed in units of frames, but for efficient processing, it is preferable to process a plurality of frames constituting one packet in units of processing. The filtered packet data is recorded in the output buffer 18. When the present invention is implemented through a computer device, it is preferable that a plurality of output buffers 18 be provided. The signals processed through the time-base conversion step and the differential amplification step are written to the plurality of output buffers 18, but the output buffers for writing the packet data are cyclically replaced. For example, assuming that there are four output buffers Buf_1, Buf_2, Buf_3, and Buf_4, the order of writing the packet data is circulated in the order of 'Buf_1-> Buf_2-> Buf_3-> Buf_4-> Buf_1 ....'.

In parallel with the writing of the packet data to the output buffer 18, the packet data recorded in the plurality of output buffers is first outputted to the audio output unit 22 in the packet data of the output buffer prior to the recording order (step S20). . That is, the cyclic order of the output buffers is also the same as the above write order of packet data. The digital data input to the audio output unit 22 is converted into an analog signal and provided to the speaker 24 through post-processing necessary for audio output such as amplification processing.

3 is a flowchart showing an algorithm of a time axis conversion step (step S16). 4 is a reference diagram for explaining a principle of obtaining a time-axis conversion signal y (·) by increasing or decreasing the length of the original audio signal x (·) corresponding to the size of the playback time adjustment rate ?? set by the user. to be. For the sake of understanding in FIG. 4, the frame Fm includes 320 samples, the reproduction time adjustment rate α is 2, and the basic difference value Sa of the starting point between each frame of the original signal x (·) is 120 (sample). For example, a check range k _max for detecting an optimal correlation based on the similarity of a waveform between a signal x (·) and a time axis conversion signal y (·) is ± 40 (sample) based on the Sa value. . These values are exemplary values and can be changed to other values according to the application environment.

First, the time-axis conversion signal is read from the first memory, that is, the first frame F0 of the original audio signal x (·) from the memory 8, and the second memory (memory 8 or the internal memory of the signal processing section 10). Copy as is (y) (step S22).

Then, the value of the frame index m is set to 1 (step S24).

Thereafter, the next loop is executed until the conversion of the original signal x (·) is completed (step S36).

First, the next frame F1 should be read from the original signal x (·) and added to the time axis converted signal y (·). Here, the start position when reading the next frame F1 from the original signal x (·) is variably determined. The criterion is determined by the synchronized lag K1 with the previous frame F0 copied in the time-axis conversion signal y (·). Also, the position to be read out and added as the time axis conversion signal y (·) of the frame F1 also varies depending on the magnitude of the reproduction time adjustment rate?.

This is generally described as follows. The synchronization delay Km generally means a change value of the start position of the next frame of the original signal x (·) so as to have an optimal correlation with the time-axis conversion signal y (·). In order to calculate the synchronization delay Km, first, the synchronization delay Km is obtained using the following equation for the range between the lower limit mSa-40 and the upper limit mSa + 40 of the correlation check range. The synchronization delay K _m is used to determine, within a predetermined range, that the next frame F _m has an optimum correlation with the previous frame F _m-1 already recorded as the time-base conversion signal y (·).

<Equation 1>

<Formula 2>

However, L means the number of samples of the overlap part between adjacent frames.

Using the two equations above the ground determined the Km synchronization delay with the best correlation, reads the next frame F _m, including N samples, using the values from the original signal x (·) of the. The read start position of the next frame F _m is S _a ± K _m, which is the value obtained by subtracting the synchronization delay K _m from the frame start interval S _a at the read start position of the previous frame F _m-1 , where 0 <S _a ± K _m < N) samples will be skipped. For example, referring to FIG. 4, when K ₁ , K ₂ , and K ₃ are set to 20, -10, and 35, respectively, the read start positions of the second, third, and fourth frames F1, F2, and F3 are 140, respectively. It is the 230th and 395th samples. Of course, the number of samples in each frame is always 320, which is N. When the original signal x (·) is read in this way, a significant amount of samples are duplicated in the previous frame and the next frame. This overlapping portion decreases (in the case of α> 1) or increases (in the case of α <1) in accordance with the magnitude of the reproduction time adjustment rate α in the time axis conversion signal y (·). Note that the reading start point of each frame does not vary regularly by the product of the basic difference value Sa and the frame index m of the starting point between each frame, but irregularly according to the size of the optimal correlation Km determined using Equations 1 and 2 Is the point.

After this reading, the frame F _m is added to the time-axis conversion signal y (·). The start position of the next frame F _m added to the time-axis conversion signal y (·) is determined by the frame start interval S _a , the reproduction time adjustment rate α, and the product m αS _a of the index m of each frame. Therefore, in FIG. 4, since αSa = 2x120 = 240, additional positions of the second, third and fourth frames F1, F2, and F3 are 240, 480, and 720. In addition, the front part of the next frame F _m overlaps with the back part of the previous frame F _m-1 . The overlapping portions of both frames are weighted using Equations 3 and 4 below, and the remaining portions of the next non-overlapping frame F _m are copied as they are.

<Equation 3>

<Equation 4>

Where g (j) is a weight function. This weight function g (j) is obviously a linear function, but an exponential function may be applied.

In this manner, the original signal x (·) is read in units of frames and added to the time axis conversion signal y (·). In this process, each time one frame is processed, the value of the frame index m is increased by one (step S30). Then, it is checked whether the size of the frame index m is a multiple of the number of frames PL constituting one packet. When the frame index m reaches a multiple of PL, the data amount of the time-base conversion signal y (·) has reached the packet data amount. In this case, the converted data is subjected to low pass filtering and then recorded in the buffer unit 18 ( Steps S32 and S34).

Then, the loop composed of steps S26 to S36 is repeatedly executed until all data of the original signal x (·) is processed while checking whether the signal has been converted to the end of the original signal x (·). In this manner, time axis conversion processing according to the reproduction time adjustment rate α set by the user is performed. FIG. 7A exemplarily shows a waveform diagram of time of a normalized original signal x (·). This waveform is a waveform diagram of the time of the signal y (·) obtained by time-base conversion so that its length is doubled (that is, when the reproduction time adjustment ratio? Is 2). Comparing these waveforms, it can be seen that the original signal x (·) is approximately doubled on the time axis, but its morphological characteristics remain almost intact, and it is possible to feel good sound quality by actually listening to the reproduction sound.

Next, FIG. 5 is a flowchart for explaining an algorithm for differentially amplifying by applying amplification higher than weak sound.

The differential amplification algorithm handles samples. Therefore, this algorithm has a feature that can be executed in parallel with the time-base conversion algorithm of FIG. For example, in the time-base conversion algorithm, it is possible to read the original signal x (·) in units of frames, perform a differential amplification algorithm on each sample of the frame, and then convert it into a time-base conversion signal y (·). That's the way. As another selectable method, when the time-axis conversion signal y (·) is collected by the packet data amount, low pass filtering is performed, and before the low-pass filtering, the differential amplification of the sample unit for the time-axis conversion signal y (·) is executed. That's the way. Of course, it is impossible to perform differential amplification after low pass filtering, so it is effective to put differential amplification ahead of low pass filtering because low pass filtering has the purpose of removing noise.

The differential amplification algorithm will be described in detail with reference to FIG. 5 as follows.

The differential amplification method can be modified in various forms depending on which function is applied as the differential amplification function and whether the differential amplification object is to be distinguished according to the value of the sample data. Basically, the differential amplification function f (x _i ) should be selected from among functions in which the amplification gain is applied relatively low from the weak sound to the strong sound. A function that satisfies this requirement is a monotonically increasing function, for example, a linear function with f (x) = x or an exponential function with f (x) = x ^{1 / β} , as illustrated in FIG. . Substituting the sample data into the variables of these functions and taking the value of the function in place of the sample data, the weakness is strengthened differentially compared to the strong sound.

(B) of Figure 6 if only the sample data of a larger value than a value optionally differential amplifier based on the threshold value x _th and apply the methods to the sample data below a threshold value x _th retains its original value as it is It is shown.

6 (c) shows that the sample data having a value larger than this value based on the threshold value x _th is used to calculate a function value using the differential amplification function f (x _i ) = x _i ^{1 / β} and the value smaller than the threshold value. The sample data having a value shows a case of applying a method of replacing the original sample data x _i by calculating a function value using a differential amplification function f (x _i ) = x _i ^β .

Of course, instead of going through the procedure of comparing the sample data x _i with the threshold value x _th , it is also possible to calculate all the sample data as a differential amplification function and convert it into the function value.

5 is a flowchart of an execution algorithm in the case of performing differential amplification according to FIG. 6C among these methods.

Referring to this flowchart, first, the size of the sample to be processed x _i is compared with a threshold value x _th (step S40). This threshold x _th can be set by the user or predefined by the system.

As a result of the comparison, when the sample data x _i is larger than the threshold value x _th , the sample data x _i is replaced by the function value x _i ^{1 / β} , and conversely, when the sample data x _i is smaller than the threshold value x _th , the sample data x _{i is used.} Is replaced by the function value x _i ^β (steps S42 and S44). If x _i is smaller than the threshold x _th , applying the exponential function to the differential amplification function as in step S42 has the advantage that the noise due to a small negative value can be effectively attenuated. On the other hand, when the linear function is applied as the differential amplification function as in the case of FIG. 6 (b), the amount of calculation is relatively small. If the original signal x (·) was recorded in a noise-free room such as a recording studio, it would not be necessary to apply an exponential function if x _i was less than the threshold x _th .

After each sample data is processed as above, it is checked whether the sample index i exceeds an integer multiple of the sample number PS of the packet data (step S46), and if the number of processed sample data has not yet reached one packet, the sample index is Only i is increased by 1 (step S48), and the operation using the new differential amplification function is repeated for the next sample data.

When the number of processed sample data reaches the packet data amount, the low-pass filtering process for noise reduction is performed on the differentially amplified data set {x _i } (step S50). Then, the filtered data is recorded in the buffer unit 18, and finally, the analog signal is converted through the audio output unit 22 and output to the speaker 24 (step S52). Of course, when the differential amplification algorithm is arranged as part of the time-base transformation algorithm, the calculation step of the sample data using the differential amplification function is only incorporated.

This processing routine is repeated repeatedly until all the sample data of the original signal is processed (steps S54 and S48).

FIG. 7C is a waveform diagram of a signal obtained by actually amplifying the original signal x (·) of FIG. 7A by applying a predetermined differential amplification factor. By comparing these two waveform diagrams, it can be seen that the low amplitude parts, rather than the loud intensity parts, are relatively more amplified. This is consistent with the intended result.

In the above, the signal processing method for increasing or decreasing the reproduction time of the digital audio signal and the signal processing method for further strengthening the weak sound have been described. Both of these methods can be implemented in parallel and therefore have several advantages. For example, if the audio signal processing method of the present invention is applied to a device for learning foreign language listening ability, the user can not only adjust the reproduction speed of the audio signal according to his or her listening ability, but also can accurately recognize the inaudible sound. The effect is expected to be greatly improved.

The audio signal processing method according to the present invention may be widely applied to a computer, an audio cassette tape recorder, as well as other audio signal reproducing apparatus, and may be used as a listening power assistance means of a user with low listening power.

In addition, the present invention provides the processing for the time-base conversion of the audio signal in the frame unit, but the audio output unit with the speaker provides a plurality of frames in a single packet unit, and the data output to the audio output unit using a plurality of output buffers The audio signal is smoothly supplied to the speaker by supplying it. As a result, the reproduction sound is not cut off and the sound quality is very good.

Although the present invention has been described above according to an embodiment of the present invention, it will be apparent to those skilled in the art that various modifications may be made without departing from the spirit of the present invention.

Claims (16)

In the method for processing a digital audio signal expressing the intensity of the sound with respect to time as an integer stream, A playback time adjustment rate α and a differential amplification rate β, which are input values designated by a user, are read from the user interface, wherein the reproduction time adjustment rate α is for adjusting the reproduction time of the audio signal longer or shorter than the normal reproduction time. a predetermined value reading step determined in a range of _min ≦ α ≦ α _max , wherein the differential amplification ratio β is for reinforcing a weak sound of an audio signal relatively more than a loud sound, and setting a predetermined value within a range of β _min ≦ β ≦ β _max ; The original audio signal x (·) stored in the first memory is continuously read in units of N samples, and the next frame F _m is smaller than the N at the start of the previous frame F _m-1 . A data reading step of reading a predetermined number of samples in duplicate between adjacent frames by reading from a sample of a position where a sample of the skipped position is skipped; The next frame F _{m is added} to the second memory so as to overlap the previous frame F _m-1 already recorded as the time-base conversion signal y (·) by a predetermined number of samples, and the samples of the overlapped portion are weighted, added, and not duplicated. The portion is added as it is, and the number of samples of the overlapped portion is added in such a manner as to increase or decrease in correspondence with the magnitude of the reproduction time adjustment rate α so that the time axis for obtaining a signal y (·) whose time axis length is converted in proportion to the reproduction time adjustment rate α is obtained. Conversion step; And Sample data x _i of each frame is amplified by substituting a predetermined differential amplification function f (x _i ) corresponding to the differential amplification ratio β, in particular, and the differential amplification function f (x _i ) is amplified from weak to strong. Including a differential amplification step where the gain is a relatively low applied function that enhances the weakness differentially compared to strong sound, And processing the reproduction time and gain control by differential amplification in parallel with the digital audio signal according to a user-specified value. The audio signal processing method according to claim 1, further comprising a normalization step of normalizing sample data of each frame read in the data reading step. The method of claim 1, wherein a plurality of output buffers are provided, and a plurality of frames of a signal processed through the time-base conversion step and the differential amplification step are bundled into one packet and written in one of the plurality of output buffers. In particular, the output buffer for writing the packet data is cyclically replaced, and in parallel, the packet data recorded in the plurality of output buffers is output such that the packet data of the output buffer prior to the recording order is first output to the audio output unit including the speaker. The audio signal processing method further comprises a data output step of controlling. The audio signal processing method according to claim 1, further comprising a filtering step of removing noise by performing low pass filtering on the differentially amplified sample data for each predetermined data amount. The read start position of the next frame F _m in relation to the data reading step is _a synchronous delay to the frame start interval S _a at the read start position of the previous frame F _m-1 . value of the acceleration a K _m S _a ± K _{m (stage, 0 <S a ± K m} <N) is the position skipped samples, the synchronization delay K _m is the next frame F within a predetermined range of the _m is the And an optimum correlation with the previous frame F _m-1 already recorded as the time-axis conversion signal y (·). The start position of each frame added to the time axis conversion signal y (·) in the time axis conversion step is the frame start interval S _a , the reproduction time adjustment rate α, and the An audio signal processing method, characterized by a product mαS _a of the index m of each frame. The audio signal processing method according to any one of claims 1 to 4, wherein the differential amplification function f (x _i ) is a monotonically increasing function. 8. The audio signal processing method according to claim 7, wherein the differential amplification function f (x _i ) is a function expressed by f (x _i ) = x _i ^{1 / β} . The differential amplification operation using the differential amplification function f (x _i ) of the differential amplifying step is read in the data reading step prior to the time-base transform processing in the time-base converting step. An audio signal processing method, characterized in that performed for each frame. The differential amplification operation using the differential amplification function of the differential amplification step is performed on the time axis transform signal y (·) obtained after the time axis transform processing in the time axis transform step. Audio signal processing method characterized in that. 5. The audio signal processing method according to any one of claims 1 to 4, wherein the reproduction time adjustment rate alpha has a value of 0.7 or more and 3 or less, and the differential amplification factor beta has a value of 1 or more and 3 or less. In the method for processing a digital audio signal expressing the intensity of the sound with respect to time as an integer stream, Reading a value of the differential amplification ratio β (where 1 <β≤3) specified by the user from the user interface; And Comparing the size of each sample data x _i of the integer stream with a threshold value x _th of a predetermined size; and The comparison, the sample data of the sample data x _i is greater than the chin value x _th of a given size, each of the samples by replacing the value of a data x _i to x _i ^{1 / β,} a value greater than the threshold value x _th And a differential amplifying step for reinforcing the weak sound relative to the strong sound for the field. (delete) The method of claim 12, wherein the audio signal processing method, the comparison of the results, each of the sample data x _i is less than the threshold value x _th the sample data x _i is more steps are replaced by x _i ^{1 / β} Audio signal processing method comprising the. The audio signal processing method according to claim 12, further comprising: performing low pass filtering on the packet data when the differentially amplified sample data is collected by the packet data amount. (delete)