CN1215875A - Pitch raising and lowering device and method for digital audio signal - Google Patents

Abstract

A tone-lifting device for digital audio signal is used to receive a digital audio signal and convert the frequency of the digital audio signal. The device mainly comprises an input device for receiving digital audio signals, a tone processing device for selecting samples with specific length in the digital audio signals to perform tone lifting processing and obtain a frequency conversion sound frame, and an audio signal connecting device for connecting the frequency conversion sound frame and the frequency conversion audio signals so as to obtain new frequency conversion audio signals. The audio signal linking device comprises a searching area potential comparing device, a linking area potential comparing device, a bit processing device and a linking device.

Description

Pitch raising and lowering device and method for digital audio signal

The present invention relates to a device and method for pitch shifting of digital audio signals, and in particular to a device for fast searching in a pitch shifting device of an unlimited audio frame division method. ) The method of connecting points between two sound frames.

In the pitch raising and lowering processing of digital audio signals, the simplest method is to speed up (pitch up) or slow down (pitch down) the playback frequency, just like speeding up or slowing down the rotation speed of a turntable. However, doing so changes the playing time length of the original audio signal. Therefore, how to keep the playing time of the original audio signal and at the same time achieve the effect of pitch raising and lowering has become the primary issue in the pitch raising and lowering processing.

Mr. Chen Siping proposed an indeterminate frame in his master thesis "On Audio Processing For MPEG Decoding, Pitch-shifting and Sub-band Coding" The implementation of the segmentation method is roughly as follows:

1. Firstly, a sound frame (audioframe) with a sample number N is taken out from the original audio signal;

2. Perform pitch up-down processing on the samples in the above-mentioned sound frame;

3. Assuming that the playback length of the sound frame after the frequency reduction becomes mN (m>1 is a falling sound; m<1 is a rising sound), then the next sound frame is determined to start from the mNth sample point in the original sound frame N samples;

4. Carry out the pitch raising and lowering processing described in step 2 for each sample in the new sound frame;

5. Find the optimal connecting point (optimum connecting point) to connect the above two sound frames;

6. Assuming that the number of samples of the sound frame after concatenation is 2mN-X, and X is the number of samples discarded due to concatenation, then again a sound frame is N samples starting from the 2mN-X sample point in the original audio signal; as well as

7. Repeat step 4 to step 6 until the pitch up and down process ends.

In the pitch raising and lowering processing of this infinitum frame segmentation method, the search and connection method of the joint point in step five is to use the last segment data of the previous frame (the so-called search region (search region) in the following text) and the next tone The data in the previous section of the box (that is, the so-called cross region in the following text) is numerically compared (the mean absolute error (MAE) is used as the basis for judging the best connection point), so as to obtain the most similar connection method. The reference equation is as follows: $\mathrm{MAE}\left(i\right)=\underset{j=0}{\overset{m-1}{\mathrm{\&Sigma;}}}|C\left(j\right)-S(i+j)|i=0~N-m$

Among them, C is the sample in the connection area, the number of samples is M; S is the sample in the search area, the number of samples is N and N>M: and the best connection point is the sample that can make the MAE have the minimum value, that is, the sample in the connection area and the search area The location of the most similar area samples.

As for the linking action, it is carried out according to the following formula: $P(i+j)=\frac{j}{m}*C\left(j\right)+\frac{m-j}{m}*S(i+j)j=0~m-1$

Among them, i is the position of the best cohesion point: P is the cohesive sample, and after P is the sample of the next frame.

Please refer to FIG. 1 , which is a schematic diagram of the pitch raising and lowering processing of the digital audio signal of the infinitum segmentation method in the case of pitch reduction.

In this illustration, it is assumed that the original audio signal S0 is a set of digital audio signals composed of a plurality of continuous digital signals. First, a group of sound frames D1 with a time length L1 is selected from the original audio signal S0 , as samples between 0 and L1-1 are marked in the figure. Then, the tone down processing is performed on the sound frame D1 (such as changing its playback speed), and the playback time becomes longer as L2, so as to obtain a set of down-frequency audio signals D1' with a time length of L2.

Next, take the distance L2 from the starting point of the previous sound frame D1 in the original audio signal S0 (that is, the time marked 0) (that is, the time length of the sound frame after frequency reduction), and then select a group of time length L1 Sound frame D2, the samples between L1 and L1+L2-1 are marked in the figure. In the same way, the pitch-down processing is performed on the sound frame D2 to obtain a set of down-frequency sound frames D2' with a time length L2.

Next, the splicing operation of the down-converted audio signals D1' and D2' is performed. Firstly, some samples of the back section of the down-converted audio signal D1' and some samples adjacent to the back section of the original audio signal S0 are selected as the search area Sa, and some samples of the front section of the down-frequency audio frame D2' are selected as the connection area Ca. Then, compare the samples in the connection area Ca and the search area Sa, and obtain the optimal connection point K1 between the down-frequency sound frames D1' and D2' according to the above mathematical formula, so as to complete the operation of the connection and obtain a down-frequency audio signal S0 '.

Then, the above steps are repeated until the whole down-frequency processing ends.

As for FIG. 2 , it shows a schematic diagram of the pitch raising and lowering processing of the digital audio signal of the infinitum method in the case of raising the pitch.

In the illustration, it is assumed that the original audio signal S1 is a set of digital audio signals composed of a plurality of continuous digital signals. First, a group of sound frames D3 with a time length L3 is selected from the original audio signal S1, as samples between 0 and L3-1 are marked in the figure. Then, the pitch-up processing is performed on the sound frame D3 (such as speeding up its playing speed), and the playing time is shortened to L4, so as to obtain a set of up-frequency audio signals D3' with a time length of L4.

Next, use the distance L4 from the starting point of the last sound frame D3 in the original audio signal S1 (i.e. the time marked 0) (i.e. the time length of the sound frame after up-conversion), and then select a group of sounds with a time length L3 Box D4, the samples between L3 and L3+L4-1 are marked in the figure. In the same way, the tone-up processing is performed on the sound frame D4, so as to obtain a set of up-frequency sound frame D4' with the number of samples L4.

Next, the splicing operation of the up-converted audio signal D3' and the up-converted audio frame D4' is performed.

Firstly, select some samples of the rear part of the up-converted audio signal D3' and some samples adjacent to the rear part of the original audio signal S1 as the search area Sb, and select some samples of the front part of the up-converted sound frame D4' as the connection area Cb. Then, compare the samples in the connection area Cb and the search area Sb, and obtain the optimal connection point K2 between the up-conversion audio signal D3' and the up-conversion sound frame D4' according to the above mathematical formula, so as to complete the operation of connection and obtain a liter frequency audio signal S1'.

And, repeat the above steps until the upscaling process is completed.

However, in the case of N=160, M=80, and the playback audio frequency is 16 kHz for the infinitum frame segmentation method, (80+79)*80=12720 addition and subtraction operations must be performed every 10 milliseconds (That is, at least 1,272,000 addition and subtraction operations are performed per second), the cost of such operations is really too high. Therefore, in order to apply the infinitum frame segmentation method to consumer products, it is necessary to find a simple and effective method for searching connection points, and the method should be completed with simple hardware circuits.

In view of this, the main purpose of the present invention is to provide a pitch raising and lowering device and method for a digital audio signal, which utilizes bit operations to simplify the search for the connection point between two sound frames, so that the calculation cost can be greatly reduced and the simple hardware can be used. line to complete.

In order to achieve the above-mentioned and other objects of the present invention, the present invention proposes a method for raising and lowering the pitch of a digital audio signal, which is used to receive a digital audio signal and convert its frequency to obtain a variable-frequency audio signal. The steps include: first, select a sample of a certain length R in the digital audio signal as the first original sound frame for frequency conversion, after frequency conversion, the time length is changed from the original L to L', so as to obtain the first sound frame of the time length L' As a frequency conversion audio signal. Then, select R samples from the time L' of the original digital audio signal as the second original sound frame for frequency conversion, so as to obtain a second frequency-converted sound frame with a new time length L'. Then, concatenate the second frequency-converted sound frame and the frequency-converted audio signal to obtain an updated frequency-converted audio signal. And, repeating the first two steps of the step, so as to complete the pitch raising and lowering of the digital audio signal.

Wherein, the aforementioned steps of connection include: firstly, select a sample of length N from the rear section of the frequency conversion signal as the search area, and compare each sample in the search area with a reference potential, so as to obtain a bit sequence of the search area, indicating the The amplitude of each sample in the search area. Then, select a sample of length M from the front section of the second frequency conversion frame as the connection area (usually N>M), and compare each sample in the connection area with the reference potential, so as to obtain a bit data of the connection area, indicating that each sample in the connection area The amplitude condition. Finally, the bit data of all sub-search areas of length M in the linking area and the search area are compared by bit comparison, so as to obtain the corresponding non-similarity; and, the linking area and the corresponding sub-search area (sub -search region) to obtain the updated frequency conversion audio signal.

In addition, in the present invention, the bit comparison of the sub-search areas of length M in the connection area and the search area can be completed by an XOR logic circuit, and the corresponding degree of dissimilarity is the number of 1s in the output result of the XOR logic circuit.

In addition, the present invention also provides a pitch raising and lowering device for a digital audio signal, which is used to receive a digital audio signal and convert its frequency. It mainly includes an input device, a tone processing device and an audio signal connecting device. The input device is used to receive digital audio signals. The pitch processing device is used to select samples of a specific length in the digital audio signal to perform pitch-lowering processing and obtain a frequency-changing sound frame. The audio signal connection device is to connect the frequency conversion sound frame with the frequency conversion audio signal, so as to obtain a new frequency conversion audio signal. Also, the audio signal connection device may include: a search area potential comparison device, which compares samples with a length of N in the back segment of the frequency conversion audio signal as the search area sample, so as to obtain bit data of a search area sample; a connection area potential comparison The device is used to compare the samples with the length M of the front section of the frequency conversion frame as the samples of the connection area, so as to obtain the bit data of a sample of the connection area; the one-bit processing device compares the bit data of the sample of the connection area with all the samples in the search area. The bit data of the sub-search area samples formed by the samples of length M, so as to obtain the corresponding dissimilarity; and a connection device, which selects the sub-search area samples with the smallest dissimilarity to perform connection with the frequency conversion sound frame and the frequency conversion audio signal.

In order to make the above-mentioned and other purposes, features, and advantages of the present invention more obvious and understandable, a preferred embodiment is specifically cited below, and in conjunction with the accompanying drawings, the detailed description is as follows. In the figure:

Fig. 1 is a schematic diagram of the pitch up and down processing of the digital audio signal of the infinitum frame segmentation method in the case of pitch down;

Fig. 2 is a schematic diagram of the pitch up and down processing of the digital audio signal of the infinitum frame segmentation method in the case of pitch up;

3A and 3B are schematic diagrams of the search area sample and the transition area sample in the pitch raising and lowering device and method of the digital audio signal of the present invention;

Fig. 4 is the circuit block diagram and the schematic diagram of the processing method thereof of the tone raising and lowering device of digital audio signal of the present invention:

FIG. 5 is a schematic diagram of pitch raising and lowering processing in the pitch raising and lowering method of a digital audio signal according to the present invention.

Continuing from the above, because the pitch-lowering process of the infinitum segmentation method uses the average absolute value difference (that is, the above mathematical formula) as the basis for the search for the connection point in the known method, it must spend a considerable amount of computing costs and The cost of hardware can only achieve the search of the connection point.

In addition, in the pitch processing of digital audio signals, since the duration of a sound frame is very small (usually between 20 and 30 milliseconds), the data in the sound frame is statistically stationary. And, because the shape, amplitude, and size of the waveforms formed by the samples in two adjacent sound frames are very similar, and the difference in pitch information often comes from the upper and lower amplitudes of the waveforms formed by the samples, therefore, the connection point can only be aimed at the search area. It is obtained by comparing the upper and lower amplitudes of the waveform composed of samples in the connecting area.

According to this characteristic, the present invention provides a device and method for pitch raising and lowering of digital audio signals in a bit-by-bit comparison mode, which only compares the up and down swings of the search area and the articulation area, so the articulation points between sound frames can be greatly reduced. The calculation and hardware cost of searching.

Please refer to FIG. 5 , which is a schematic diagram of pitch up/down processing in the pitch up/down method of a digital audio signal according to the present invention, which is used to receive a digital audio signal and convert its frequency to obtain a variable frequency audio signal.

Firstly, a sample of a specific length in the digital audio signal is selected as the first original sound frame for frequency conversion, so as to obtain a first sound frame with a time length L6 as the frequency-converted audio signal.

In the illustration, it is assumed that the original audio signal S2 is a set of digital audio signals composed of a plurality of continuous digital signals, as shown in FIG. 1 and FIG. 2 . First, a sound frame with a time length L5 (that is, samples between 0 and L5-1 marked in the figure) is selected from the original audio signal S2 as the first original sound frame D5. Then, the first original sound frame D5 is subjected to pitch up-and-down processing (such as speeding up or slowing down its playback speed), then the time of playing the sound is shortened or increased to become L6, so as to obtain the first sound frame D5' of a group of time length L6. is the desired frequency conversion audio signal S2'.

Then, select the sample of the L5 time length starting from the first original sound frame time L6 (i.e. the time length of the first sound frame after frequency conversion) in the digital audio signal (that is, the one marked L6～L5+L6-1 in the figure Samples in between) are pitched and lowered for the second original sound frame D6, so as to obtain a set of second frequency-converted sound frames D6' with a time length L6.

Then, the frequency conversion audio signal D5' and the frequency conversion sound frame D6' are connected.

Different from the existing ones, the present invention reduces the complexity of existing value comparisons by using bit comparison.

Please refer to FIG. 3A and FIG. 3B , which are respectively used to represent the schematic diagrams of the samples of the search area and the transition area in the pitch raising and lowering method of the digital audio signal of the present invention. Wherein, the search area Sc may be N samples selected from the rear segment of the frequency-converted audio signal D5' and the adjacent digital sound frame S2. The concatenation area Cc may be M samples selected from the front section of the second frequency conversion sound frame D6'.

Moreover, in this step, the reason why the search area Sc includes part of adjacent digital audio signals is to find a reasonable connection point. And through the connection of several frequency conversion sound frames, such as the connection of D5' and D6', the overall playback time after frequency conversion is closer to the original digital audio signal S2.

Also, please refer to FIG. 4 , which is a schematic diagram of the pitch processing of the pitch raising and lowering method of the digital audio signal according to the present invention. Among them, in order to save computing and hardware costs, first, the samples in the search area Sc and the connection area Cc are compared in potential through a set of potential comparison devices 20, 30, so that the samples that are greater than the reference potential Vref (such as 0V) output "1" ", and the samples that are lower than the reference potential will output "0", which respectively represent the upper and lower amplitudes of the samples in the search area Sc and the connection area Cc, and are respectively used as the search area bit data Sd and the link area bit data Cd.

Then, compare the bit data of the sub-search area Ssub in the linking area Cc and the search area Sc each having the length of the linking area Cc in a bit comparison manner. And calculate the number of different bits representing the bit data of the connection area Cc and each sub-search area Ssub in the processing result of the XOR logic circuit, and use it as the degree of dissimilarity corresponding to the connection area Cc and each sub-search area Ssub. In this embodiment, the comparison of the connecting area sample Cc and each sub-search area sample Ssub can be completed by using an XOR operation device, that is, calculating the number of bits representing 1 in the result of the XOR operation (that is, the number of different bit data) , used as its corresponding dissimilarity.

Finally, join the sub-search area corresponding to the minimum dissimilarity in the sub-search area Cc with the sub-search area Ssub (for example, use a counting device to calculate the number of 1s in the processing result of the XOR logic circuit as the dissimilarity corresponding to each sub-search area , and compare the degrees of dissimilarity with a numerical comparator to find the sub-search region corresponding to the minimum degree of dissimilarity and its corresponding sample K), so as to obtain a new frequency-converted audio signal S2'; and, repeat the above steps to Complete the entire pitch up and down process.

In this method, since the digital audio signal is subdivided into sound frames with a length of about 20-30 milliseconds, and the comparison between the connection area Cc and the sub-search area Ssub is completed by a potential comparison device, the cost of the circuit can be reduced, and the Applicable to general tone processing devices.

In addition, the present invention also provides a digital audio signal tone device, which is used to receive a digital audio signal composed of a plurality of samples sequentially, and convert it into a frequency-changed audio signal. The device mainly includes: an input device, a tone processing device and an audio signal connecting device. The input device is used to receive digital audio signals. The pitch processing device is used to select samples of a specific length in the digital audio signal to perform pitch up-down processing, so as to obtain a frequency-changed sound frame. The audio signal connection device connects the frequency conversion sound frame to the frequency conversion audio signal, so as to obtain a new frequency conversion audio signal.

Wherein, the audio signal connection device includes: a search area potential comparison device 20 , a connection area potential comparison device 30 , a bit processing unit 40 and a connection device 50 .

The search area potential comparison device 20 is to select a specific length from the rear segment of the frequency conversion audio signal S2' and its adjacent digital audio signal S2, such as N samples as the search area for potential comparison, so as to obtain a search area bit data Sd, corresponding to The amplitude of each sample in the search area Sc. Such as: select a sample of length N from the back section of the sound frame D5' and the adjacent digital audio signal S2 as the search area, and use a potential comparison device 20 to compare it with a reference potential, such as 0V, so as to obtain a sound Bit data for box D5'.

And the connection area potential comparison device 30 selects samples of length M from the front section of the second frequency conversion audio signal D6' as the connection area for potential comparison, so as to obtain a connection area bit data Cd, corresponding to the amplitude of each sample in the connection area Cc Condition. For example: use a potential comparison device 30 to compare a reference potential, such as 0V, with samples of the length M of the front section of the sound frame D6', so as to obtain the bit data of a sound frame D6'.

The bit processing device 40 is used to compare the bit data of the sub-search area Ssub formed by the samples of length M in the connection area Cc and the search area Sc, so as to obtain a corresponding degree of dissimilarity. For example, using an XOR logic circuit to sequentially compare the similarities and differences of each bit in the bit data of the sound frames D5' and D6', and calculate the number of bit "1" or "0" in the result (such as using a counter to achieve), as the corresponding degree of dissimilarity.

The concatenating device 50 finds the sub-search area corresponding to the minimum dissimilarity and the corresponding sample k, and concatenates the frequency-converted sound frame D6' to the frequency-converted audio signal D5', so as to obtain the new frequency-converted audio signal S2'. For example, compare all dissimilarities with a comparator, and obtain a sub-search area sample with the minimum dissimilarity therefrom, then, use the starting point (such as sample point K) of this sub-search area sample as the sound frame D5' and D6' The connection point is used to connect the two frequency conversion sound frames.

In summary, the pitch processing device and method of the digital audio signal of the present invention can apply a simple circuit to complete the pitch raising and lowering processing of the indeterminate frame segmentation method. Used in general tone processing systems.

Although the present invention has been disclosed as above with preferred embodiments, it is not intended to limit the present invention. Anyone skilled in the art can make changes and modifications without departing from the spirit and scope of the present invention. Therefore, the present invention The scope of protection should be defined by the appended claims.

Claims (12)

1. A method for pitch raising and lowering of a digital audio signal is used to receive a digital audio signal and convert its frequency to obtain a variable frequency audio signal. It is characterized in that the method includes the following steps: (a) Select a sample of a certain length R in the digital audio signal as the first original sound frame for frequency conversion, and the time length of the first original sound frame changes from L to L' after frequency conversion, so as to obtain a time length L' The first sound frame is used as the frequency conversion audio signal; (b) performing frequency conversion with samples of a specific length R starting at the time L' in the original digital audio signal as the second original sound frame, so as to obtain a second frequency-changed sound frame of a time length L'; (c) connecting the second frequency conversion sound frame and the frequency conversion audio signal, so as to obtain the updated frequency conversion audio signal; and (d) repeat steps (b), (c), so as to complete the pitch up and down of the digital audio signal; Wherein, step (c) comprises: Select a sample of length N from the back section of the frequency conversion signal and its adjacent digital audio signal as a search area, and compare each sample in the search area with a reference potential to obtain a search area bit data, indicating the search The amplitude of each sample in the area; Select a sample of length M from the front section of the second frequency conversion sound frame as a linking area, and compare each sample in the linking area with the reference potential to obtain bit data of the linking area, indicating the amplitude of each sample in the linking area ; Comparing the bit data of the linking area and all the sub-search areas of length M in the search area by means of bit comparison, so as to obtain the degree of dissimilarity corresponding to the linking area and the sub-search areas; and The joining area and the sub-search area corresponding to the minimum dissimilarity among the sub-search areas are joined to obtain an updated frequency-converted audio signal. 2. The pitch raising and lowering method according to claim 1, wherein, the step of forming the degree of dissimilarity corresponding to the search area and the sub-search intervals comprises: Comparing the bit data of the linking area and the sub-search areas of length M in the search area by means of bit comparison, so as to obtain bit data of dissimilarity; and Calculate the number of bits representing bit differences in the dissimilarity bit data, and use it as the dissimilarity corresponding to the linking area and the sub-search area. 3. The pitch raising and lowering method as claimed in claim 1, wherein the length N of the search region samples is greater than the length M of the transition region samples. 4. The pitch raising and lowering method as claimed in claim 3, wherein the search area samples are composed of the next N samples of the frequency conversion audio signal. 5. The pitch raising and lowering method as claimed in claim 1, wherein the step of bit-comparing the search region sample and all the sub-search buffer frames composed of consecutive samples of length M in the link region sample is completed by an XOR step. 6. The pitch raising method as claimed in claim 5, wherein the degree of dissimilarity is represented by the number of "1" bits in the result of the XOR step. 7. A pitch raising and lowering device for a digital audio signal is used to receive a digital audio signal and convert its frequency to obtain a variable frequency audio signal. It is characterized in that the pitch raising and lowering device includes: an input device for receiving the digital audio signal; a pitch processing device, which is used to select samples of a specific length in the digital audio signal to perform pitch up-down processing, so as to obtain a frequency-changed sound frame; and An audio signal connection device, which connects the frequency conversion sound frame to the frequency conversion audio signal, so as to obtain the updated frequency conversion audio signal: Wherein, the audio signal connection device includes: A search area potential comparison device, which selects a sample of length N from the rear segment of the frequency conversion audio signal and its adjacent digital audio signal as the search area for potential comparison, so as to obtain a search area bit data, indicating the value of each sample in the search area Amplitude condition; A linking area potential comparison device, which selects a sample of length M from the front section of the frequency conversion sound frame as the linking area for potential comparison, so as to obtain a linking area bit element data, indicating the amplitude of each sample in the linking area; a one-bit processing device for comparing the bit data of the sub-search area samples formed by the samples of length M in the joint area and the search area, so as to obtain the degree of dissimilarity corresponding to the joint area and the sub-search area; and A concatenating device concatenates the concatenated area with the sub-search area corresponding to the minimum dissimilarity among the sub-search areas, so as to obtain the updated frequency-converted audio signal. 8. The pitch raising and lowering device as claimed in claim 7, wherein the potential comparing means of the connection area uses a reference potential for potential comparison. 9. The pitch raising and lowering device as claimed in claim 7, wherein the search area potential comparing means uses a reference potential for potential comparison. 10. The pitch raising and lowering device according to claim 8 or 9, wherein the reference potential is 0V. 11. The pitch raising and lowering device as claimed in claim 7, wherein the bit processing device is a logic circuit for performing XOR operation. 12. The pitch raising and lowering device as claimed in claim 7, wherein the degree of dissimilarity is represented by the number of "1" bits in the result of the XOR step.

Images