CN101615397B - Audio signal processing method - Google Patents

Abstract

A method of audio signal processing, comprising the steps of: dividing an audio signal data stream into a plurality of selection sections; determining a target segment in the audio signal data stream, the target segment including a splicing point for splicing the splicing segment; selecting one of the selection sections as the connection section according to at least one parameter in the target section; and operating the target section and the connection section to connect the connection section to the connection point and output a processing section. The invention allows fast processing and achieves a high weight sound reproduction.

Description

audio signal processing method

technical field

The present invention relates to an audio signal processing method, and in particular to a variable speed audio signal processing method.

Background technique

Time scaling (time scaling) is to change the length of the audio signal data stream (that is, change the number of data points) to change its playback speed without affecting the pitch.

Referring to Fig. 1, time domain harmonic scaling (TDHS) is a commonly used time scaling technique, which has the advantage of low computation. The audio signal data stream includes a plurality of audio signal sections, and each audio signal section includes a plurality of point data. As shown in Figure 1 (a), when playing at a slower speed, find out a segment of audio signal data S2 from the audio signal segment S1 and the accompanying audio signal data, splice (splice) after the audio signal segment S1, to increase the length of the audio signal segment S1. As shown in FIG. 1( b ), when playing at a faster speed, the audio signal segment S3 and the audio signal segment S4 can be partially connected to shorten the total length of the audio signal segments S3 and S4 .

However, in the TDHS technology, if the connection point and the connection section are not selected properly, the sound will sound discontinuous, or noise will be generated, resulting in reduced sound reproduction quality. In addition, when the audio signal is played together with the video signal, no matter whether it is played at a faster speed or a slower speed, the two must be synchronized so that the sound and image will not be uncoordinated, but the TDHS technology may not be able to change the audio signal. After the length reaches the target length, this will cause the audio signal to be out of sync with the video signal.

Contents of the invention

Therefore, the object of the present invention is to provide an audio signal processing method that can process quickly and achieve high-quality sound reproduction.

Therefore, the audio signal processing method of the present invention includes the following steps: dividing the audio signal data stream into a plurality of selected sections; determining a target section in the audio signal data stream, and the target section includes a junction point for connecting the junction areas segment; according to at least one parameter in the target segment, selecting a segment from the selection segment as the concatenated segment; and operating the target segment with the concatenated segment to concatenate the concatenated segment to the join point and output the processing section.

Description of drawings

Figure 1 (a), (b) is a schematic diagram illustrating the TDHS technology;

Fig. 2 is a flowchart illustrating a preferred embodiment of the audio signal processing method of the present invention;

Figure 3 (a)-(d) is a schematic diagram illustrating the various steps of the preferred embodiment;

Figure 4 is a flowchart illustrating further steps of the preferred embodiment; and

5(a), (b) are schematic diagrams illustrating the curve fitting method used in the preferred embodiment.

Description of main component symbols

21~24 steps

30 reference points

32 articulation points

34 end points

36 starting point

38 intermediate points

41~44 steps

AS target segment

AD processing section

C1～CN selection section

CR junction

f(x) polynomial

Detailed ways

The aforementioned and other technical content, features and effects of the present invention will be clearly presented in the following detailed description of a preferred embodiment with reference to the drawings.

With reference to Fig. 2 and Fig. 3, it illustrates the preferred embodiment of audio signal processing method of the present invention, comprises the following steps:

Step 21: Divide the audio signal data stream into a plurality of selection sections C1-CN;

Step 22: Determine a target segment AS in the audio signal data stream, the target segment AS includes a connection point 32 for connecting the connection segment CR;

Step 23: According to at least one parameter of the target segment AS, select a segment from the selected segments C1-CN as the concatenated segment CR; and

Step 24: Perform an operation on the target segment AS and the concatenated segment CR to concatenate the concatenated segment CR to the concatenated point 32, and output a processed segment AD.

Can be understood by above-mentioned steps, an embodiment of the audio signal processing method of the present invention is to divide audio signal data flow into a plurality of selection sections C1～CN (step 21) earlier, as shown in Figure 3 (a); And in Find the target segment AS in the audio signal data stream for audio signal processing, wherein the target segment AS includes a junction point 32 for joining the junction segment CR (step 22), as shown in Figures 3(b) and 3(c) shown. According to the masking effect of sound psychology, the junction point 32 can be selected after the reference point 30 with a larger amplitude, and according to an embodiment, the amplitude value of the reference point 30 is greater than the average amplitude value of the audio signal data stream. In order to find the most appropriate connecting segment CR among the selected segments C1-CN, the embodiment of the present invention firstly calculates the maximum amplitude, minimum amplitude and Zero crossing rate (zero crossing rate), and then, in the selected section C1~CN, find out that the maximum amplitude, minimum amplitude and zero crossing rate are closest to the maximum amplitude, minimum amplitude and zero crossing rate in the target section AS Or, as the concatenated segment CR (step 23). The maximum amplitude and the minimum amplitude are used to judge the volume, so as to avoid the volume difference between the target section AS and the connecting section CR. It is a male voice, or the target section AS is a male voice and the connecting section CR is a female voice. It should be noted that referring to the maximum amplitude, minimum amplitude and zero-crossing rate as the basis for selection is only an embodiment. The present invention is not limited thereto, and other parameters as the basis for selection. After the linking section CR is found, the present invention performs a linear or non-linear process between the partial data of the target section AS (from the linking point 32 to the end point 34) and the partial data of the linking section CR (from the starting point 36 to the middle point 38). Linear superposition operation (overlap and add, OLA) to output processing section AD (step 24), as shown in Fig. 3(d).

With reference to Fig. 4, in order to make audio signal and video signal synchronous, the preferred embodiment of audio signal processing method of the present invention also comprises the following steps:

Step 41: Calculate the error value=number of target audio signal data points-the number of audio signal data points in the processing section AD, and express the error value in fractional form q/p;

Step 42: Determine the size of q/p, if q/p>0, then skip to step 43, if q/p<0, then skip to step 44, otherwise end;

Step 43: adding a total of q points of audio signal data in every consecutive p processing sections AD; and

Step 44: Totally reduce q points of audio signal data in every consecutive p processing sections AD.

For example, if the number of data points of the audio signal in the processing section AD=258, and the number of data points of the target audio signal=258.2, then the error value=0.2, p=5, q=1 can be selected, and in every 5 consecutive processing areas In section AD, make the number of audio signal data points of 4 processing sections AD=258, and the number of audio signal data points of 1 processing section=259. In this way, the average number of audio signal data points of these 5 processing sections=258.2 .

There are two ways how to increase or decrease the audio signal data in the processing section AD:

(1) Curve fitting method (curve fitting method)

When the u-point audio signal data will be added in the processing section AD, first arbitrarily select continuous v-point audio signal data from the processing section AD, and obtain the h-order polynomial f( x), and then obtain equidistant v+u point audio signal data according to the polynomial f(x) to replace the v point audio signal data, wherein, v>u, h≥v.

With reference to Fig. 5 (a), for example, when will increase 1 point audio signal data in processing section AD, can select continuous 4 point audio signal data arbitrarily from processing section AD earlier, and find out that can represent The 4th order polynomial f(x)=b ₄ x ⁴ +b ₃ x ³ +b ₂ x ² +b ₁ x+b ₀ of these 4 audio signal data, and then calculate the equidistant according to this polynomial f(x) The 5 points of audio signal data to replace the 4 points of audio signal data.

Similarly, when the u-point audio signal data is to be reduced in the processing section AD, the continuous v-point audio signal data is arbitrarily selected from the processing section AD, and the h order that can represent the v-point audio signal data is obtained. polynomial f(x), and then calculate equidistant v-u point audio signal data according to the polynomial f(x) to replace the v point audio signal data, wherein, v>u, h≥v.

With reference to Fig. 5 (b), for example, when wanting to reduce 1 audio signal data in processing section AD, can first select arbitrarily continuous 4 audio signal data from processing section AD, and find out can represent The 4th order polynomial f(x)=b ₄ x ⁴ +b ₃ x ³ +b ₂ x ² +b ₁ x+b ₀ of these 4 audio signal data, and then calculate the equidistant according to this polynomial f(x) 3 points of audio signal data to replace the 4 points of audio signal data.

(2) TDHS method

When the u-point audio signal data will be added in the processing section AD, the u-point audio signal data immediately after the joining section CR can be added to the end of the processing section AD, and when it is to be decreased in the processing section AD When the u-point audio signal data is used, the last u-point audio signal data in the processing section AD is discarded.

Preferably, in step 43, the first k processing sections AD are increased by nint(kq/p) point audio signal data in total, and in step 44, the previous k processing sections AD are decreased by |nint(kq/p p)|point audio signal data, so that the average audio signal data point number of the processing section AD reaches the maximum accuracy, wherein, nint(x) represents the integer closest to x, and if there are two integers closest to x, Choose one of them arbitrarily.

In this embodiment, in step 43, the number of audio signal data points w _i added in the i-th processing section is as follows:

${\mathrm{<span\; class="notranslate">\; w</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; =</span>\left\{\begin{array}{c}\mathrm{<span\; class="notranslate">\; nint</span>}<span\; class="notranslate">\; (</span>\frac{\mathrm{<span\; class="notranslate">\; q</span>}}{\mathrm{<span\; class="notranslate">\; p</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}\\ \mathrm{<span\; class="notranslate">\; nint</span>}<span\; class="notranslate">\; (</span>\frac{\mathrm{<span\; class="notranslate">\; iq</span>}}{\mathrm{<span\; class="notranslate">\; p</span>}}<span\; class="notranslate">\; -</span>\underset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; w</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ></span>\mathrm{<span\; class="notranslate">\; 1</span>}\end{array}\right.$

And in step 44, the number of audio signal data points w _i reduced in the i-th processing section is as follows:

${\mathrm{<span\; class="notranslate">\; w</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; =</span>\left\{\begin{array}{c}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; nint</span>}<span\; class="notranslate">\; (</span>\frac{\mathrm{<span\; class="notranslate">\; q</span>}}{\mathrm{<span\; class="notranslate">\; p</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; |</span><span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}\\ <span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; nint</span>}<span\; class="notranslate">\; (</span>\frac{\mathrm{<span\; class="notranslate">\; iq</span>}}{\mathrm{<span\; class="notranslate">\; p</span>}}<span\; class="notranslate">\; +</span>\underset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; w</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; |</span><span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ></span>\mathrm{<span\; class="notranslate">\; 1</span>}\end{array}\right.$

For example, when p/q=1/5, 2/5, -1/5 or -2/5, the number of audio signal data points w _i increased or decreased in the i-th processing section is shown in the table below .

p/q=1/5 p/q=-1/5 p/q=-1/5 p/q=-1/5 w ₁ 0 0 0 0 w ₂ 0 1 0 1 w ₃ 1 0 1 0 w ₄ 0 1 0 1 w ₅ 0 0 0 0

In summary, the present invention finds the connecting point 32 after the point with a large amplitude (ie, the reference point 30), and searches for the connecting section CR through the maximum amplitude, minimum amplitude and zero-crossing rate, so that the processing speed is faster, and The sound reproduction quality is high, and the audio signal can be synchronized with the video signal by making the average audio signal data point number of the processing section AD reach the target audio signal data point number.

The above content is only a preferred embodiment of the present invention, and the scope of the present invention cannot be limited with this, that is, simple equivalent changes and modifications made according to the scope of the claims of the present invention and the contents of the description of the invention still belong to the scope of the present invention Inside.

Claims (12)

1. acoustic signal processing method includes:

Audio signal data is divided into a plurality of selection sections;

Determine the target section in this audio signal data, this target section comprises and is connected point in order to be connected section;

According at least one parameter in this target section, select to select the section section to be connected section as this from this; And

This target section is connected section carries out computing and be attached to this linking point should be connected section with this, and section is processed in output.

2. acoustic signal processing method according to claim 1 wherein, selects the step of this linking section also to select the second parameter in the section to decide this linking section according to each.

3. acoustic signal processing method according to claim 2, wherein, the second parameter and this parameter in this target section of selecting the step of this linking section to seek in this selection section are immediate, to determine this linking section.

4. acoustic signal processing method according to claim 1, wherein, this parameter is at least one in peak swing, minimum amplitude and the zero-crossing rate of this target section.

5. acoustic signal processing method according to claim 1, wherein, this linking point is positioned at after the reference point, and the amplitude of this reference point is greater than the mean value of amplitude of this audio signal data stream.

6. acoustic signal processing method according to claim 1 wherein, is connected section with this target section and carries out computing and carry out the superposition computing for this target section is connected section with this with this.

7. acoustic signal processing method according to claim 1 also comprises following steps:

Error of calculation value, this error amount are that the audio signal data that the target audio signal number of data points deducts this processing section is counted, and this error amount is represented with mark pattern q/p;

When q/p＞0, altogether increasing q point audio signal data in p processing section continuously every; And

When q/p＜0, altogether reducing q point audio signal data in p processing section continuously every.

8. acoustic signal processing method according to claim 7, wherein, the use curve fitting process changes the audio signal data of this processing section and counts.

9. acoustic signal processing method according to claim 8, wherein, the mode that increases u point audio signal data in this processing section is:

From this processing section, choose continuous v point audio signal data;

Obtain the h rank polynomial expression that can represent this v point audio signal data; And

Obtain equidistant v+u point audio signal data according to this polynomial expression, to replace this v point audio signal data;

Wherein, v＞u, h 〉=v.

10. acoustic signal processing method according to claim 8, wherein, the mode that reduces u point audio signal data in this processing section is:

From this processing section, choose continuous v point audio signal data;

Obtain the h rank polynomial expression that can represent this v point audio signal data; And

Obtain equidistant v-u point audio signal data according to this polynomial expression, to replace this v point audio signal data;

Wherein, v＞u, h 〉=v.

11. acoustic signal processing method according to claim 7, wherein, the mode that increases u point audio signal data in this processing section is: will be close to this linking section u point audio signal data afterwards and be added in this processing section backmost.

12. acoustic signal processing method according to claim 7, wherein, the mode that reduces u point audio signal data in this processing section is: the last u point audio signal data that will process section is given up.

Images