JPH0366000A - audio playback device - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

[Detailed description of the invention] Industrial field of application Invention 41 An audio recording and reproducing device for recording and reproducing voices and various acoustic signals, and an audio recording and reproducing device that is installed in the audio recording and reproducing device to change the speech rate during conversation, etc. A return to conventional technology regarding audio playback speed conversion methods With advances in semiconductor technology, equipment that converts audio and audio signals into digital signals and records and plays them in semiconductor memory has been put into practical use. Although there has been a lot of interest in audio playback speed conversion methods and audio recording/playback devices that allow recorded audio to be played back at a speed different from the speed at which it was recorded while maintaining its original pitch, the following will explain conventional audio recording/playback devices. FIG. 7 shows the configuration of a conventional audio recording and reproducing device, in which 11 is an AD converter, 12 is a DA converter, and 13
14 is a memory; 14 is a data writing device; 15 is a data reading device; 16 is a muting device; and 17 is a speed control device. The data is converted into a digital signal by the AD conversion device 11 and written to the memory 13 by the data writing device 14. The speed control device 17 controls the data reading device 15 according to the speed conversion ratio, and the data is read from the memory 13 by the data reading device 15. Read data. Depending on this data reading method, the playback speed can be varied in various ways. When increasing the playback speed, the data read out in block units is thinned out. When the playback speed is slowed down, the data read out in block units is repeated. Discontinuous parts between each block are removed using the muting device 16. As an example, cases where the speed conversion ratio is 0.5 and 2.0 are schematically shown in FIG. 8. Here, the speed conversion ratio is defined by the following equation.

Problems to be Solved by the Invention In the above conventional configuration, data is thinned out or repeated in order to change the speed. However, when increasing the speed, there is some data that is not used, and the intelligibility of consonants etc. deteriorates, but discontinuities in the signal waveform also occur, and if these parts are muted, the sound is played back with a lack of naturalness. In particular, the method using the pitch period cannot be applied to music for which pitch extraction is difficult. An object of the present invention is to provide an audio recording and reproducing device that is rich in naturalness and does not cause omissions, and a method for converting audio playback speed. a buffer memory for temporarily storing the data, a weighted addition calculating means for reading out the data on the buffer memory, adding weighted sums to two pieces of data separated by a predetermined length of time, and writing the result to the buffer memory; The present invention is characterized by comprising speed control means for controlling weighted addition calculations and the like.

Effect: With this configuration, the weighted addition calculation means performs a weighted addition calculation on the data temporarily recorded in the buffer memory to cancel discontinuity, and by multiple reading data from this buffer memory, Embodiment An embodiment of the present invention will be described below with reference to the drawings. Fig. 1 shows an audio recording and reproducing apparatus according to an embodiment of the present invention. In Figure 1, 11 is an AD converter fk, 12 is a DA converter, and 13 is a memo for one person.
4 is a data writing device, which is similar to the conventional example & 2
1 is a data transfer device; 22 is a buffer memory; 23 is a weighted addition calculation device; 24 is a data reading device; and 25 is a speed control device.

The operation of the audio recording and reproducing device of this embodiment configured as described above will be explained below. Although the operation is the same as that of the conventional example until the input audio signal is recorded in the memory 13, the data transfer device 21 has a high speed. A predetermined number of data is transferred to the buffer memory 22 in block units at a constant cycle under the control of the control device 25. This transferred data is weighted and averaged by the weighted addition calculation device 23 under the control of the speed control device 25. , even if the data is written to the buffer memory 22 again, the data reading device 24 reads data at a predetermined address at a constant cycle under the control of the speed control device 25.
The control of the speed control device changes the process according to the speed conversion ratio. A schematic diagram of the speed conversion algorithm in the case of .3 is shown. In Figure 2, we can see how the audio data changes over time.A Processing algorithm , (c) 1<α<α2, (d) 2≦α, and the third is the fourth is the fifth.
FIG. 6 shows a flowchart of the data control procedure in the cases of ((), (+1), (8), and (d), respectively).

Here, the address of the audio signal data (unprocessed speed conversion) to be read from the memory 13 by the data transfer device 21 is set to IA,
The address of the data read out from the buffer memory 22 by the data reading device 24 after speed conversion processing is IA, the segment length to be weighted and added is Tx, the value of the data in the memory 13 existing at address An is M(An), and the value existing at address Ak is B M (A
k), the weighting function is w(n), and the address of the memory 13 indicating the end of the audio file to be played is IEO.
The flowchart of FIG. 3 (in the case of α≦0.5) will be explained below. (a) First, calculate the number of speed conversion output data T0 that is output per cycle of processing. (b) Next, initialize IA and IB.
(C) The initial value of IB shall be set considering that data will not be overwritten in the following processing.

(C) Next, 27a pieces of data are transferred from the address indicated by IA on the memory 13 to the buffer memory 22. (d) Next, from -0 to (Tx -1), weighted with window function w(k) as shown below, and then added and recorded from address 1113 onward.8M(IB+k)1w(k)・BM(k)+( 1-w(
k))・BM(T-+k).

(0≦≦Tx-1) (Formula 3.l) Here, w(k) is a window function that monotonically decreases with respect to k. For example, in a triangular window, it can be expressed as the following equation. 2Tx each for value IA and IB
, Tx is added (f) Next, if the value obtained by adding 2Tx to the address value IA is greater than or equal to the address IEOF of the memory 13, a branch is made to terminate the processing (\ Otherwise, (g ) Continue the following processing.

(g) Next, similarly to the process in (d), the calculation expressed by the following equation is performed from -Tx to Tx-1.

BM(k-T-)1w(k)-BM(k)+(1-w(
k))・BM(T-+k).

(Tx≦≦Tx−1) (Equation 3.3) (h) Then −
The data in the memory 13 from 0 to 2Tx -Tx -1 is transferred to a predetermined address in the buffer memory 22 to complete one cycle of processing, and (d) the following processing is repeated.

With the above one-cycle processing, the output data To is obtained for the input data Ts + To, and the processed data (Yo) is stored consecutively starting from the IB address initialized in (b) of the buffer memory 22. Below is the fourth
The flowchart shown in the figure (in the case of 0,5<α×1) will be explained. (a) First, calculate the number of data Tx that will be output as is in one cycle of processing, (b) Next, initialize IA and IB.
(C) The initial value of IB shall be set considering that data will not be overwritten in the following processing.

(C) Next, 27. from the address indicated by IA on the memory 13. After transferring the data to the buffer memory 22, (d) Next, the data on the buffer 7 memory 22 that is Tx apart is divided by Tx from -0 to (Tx-1') using a window as shown in the following equation. After weighting with the function w(k), add it and record it after address IB.
k))・BM(Tx+k).

(O≦to≦Tx-1) (Equation 4.1) Here, w(k) is the same as in Figure 3, but (e) then from =o to T8-1 as shown below. Transfer data from memory 13 to buffer memory 22 By (IB+Tx+k) 4-M (IA+27@+k)
, (0≦≦Tx-1) (Formula 4.2) (f) Next, add 27s to the address value I and IB, respectively.
Even if TX, Ts + TI are added, one cycle of processing is completed after (g) or more, so the value obtained by adding 27s + Tx to the address value IA is the memory 1.
If it is greater than or equal to the address IEOF of 3, take a branch that ends the process to line t\ Otherwise, (
c) Continue the following processing.

With the above l-cycle processing, input data 27a 10Tx
Even if the output data of Ts +TX is obtained for
Processed data (even if the processed data is stored consecutively after the IB address initialized in (b) of the buffer memory 22) is explained below. a) First, calculate the number of data Tx that will be output as is in one cycle of processing. (b) Next, initialize IA and IB.
(c) The initial value of IB shall be set considering that the data will not be overwritten in the following processing.
C) Next, 2 from the address indicated by IA on the memory 13
Transfer T data onto the buffer memory 22 (d) Next, transfer data from -0 to Tx-1 within the buffer memory 22 as shown below: 8M(IB+k) 18M(k) , (0≦to≦T
11) (Formula 5.1) (e) Next, for the T-separated data on the buffer memory 22, apply a window function w(k) from -0 to (Tx-1) by T1 as shown in the following equation. After weighting with
)+w(k)・BM(Tx+k).

(0≦≦Tx-1) (Equation 5.2) Here, w(k) is the same as in Figure 3, but (f) then from -0 to Tx-1 as shown below. Data is transferred within the buffer 7 memory 22.

BM (IB+27・10k) 18M (Tx 10k),
(0≦to≦Tx-1) (Formula 5.3> (g) Next, data is transferred from the memory 13 to the buffer memory 22 from -T to Tx-1 as shown below. )1M (IA+T・10k)
, (To≦≦Tx−1) (Formula 5.4) (h) Next, add Ts+Tx to address value I and IB respectively.
, 2T.

+Tw is added and one cycle of processing is completed with (g) or more. Line t, X, otherwise (e)
Continue the following processing.

For the above l-cycle processing type human data Ta +Tx, output data of 27s +TX can be obtained.
The flowchart of FIG. 6 (in the case of α≧2) will be explained below. (a) First, calculate the number of input data Tx consumed in one cycle of processing (b) Next, IA,
Initialize IB. However, (C) Considering that the data will not be overridden in the following processing, IB
The initial value of is assumed to be set.

(C) Next, transfer 2Tx data from the address indicated by IA on the memory 13 onto the buffer memory 22. (d) Next, transfer data from -0 to Tx-1 in the buffer memory 22 as shown below. BM (IB 10k) 18M (k) , (0≦to≦T
11) (26.1) (e) Next, for the T-separated data on the buffer memory 22, apply the window function w(k ) and then add it to record after address IB+T- BM(IB+Tx+k)←(1-w(k))・BM(k
)+w(k)-BM(Tx+k).

(0≦≦Tx-1) (26.2) Here, w(k) is the same as in Figure 3, but (f) next, address values IA and IB are 2Tx and Ts+T, respectively.
.

(g) Next, address value IA becomes address IE of memory 13.
If it is greater than or equal to OF, take a branch to end the process (\) Otherwise, (h) continue the following process.

(h) Next, in the same way as in (e), calculate the following formula from -T to (Tx-1>) and record after address O BM
(k-T・)←(1-w(k))・BM(k>+w(k
)・BM (TI 10k).

(Tx≦≦Tx-1) (26.3) (i) Next -
Buffer 7 memory 2 as shown below from 0 to Tx-1
Transfer data within 2, but BM (Tx-Tx+k) 18M (T・+k), (0≦≦Tx-1) (26.4) (j) Next, from =o to Tx-1 and below Transfer the data from the memory 13 to the buffer memory 22 as shown in
End cycle processing L (d) Repeat the following processing.

BM (2Tx-T・+k) 1M (IA 10k),
(0≦to≦T11) (26.5) Processing type of the above l cycle T for input data Tx
Even if the output data of s + T@ is obtained, and even if the processed data (above) is stored consecutively after the IB address initialized in (b) of the buffer memory 22, the weight function w(n) is shown in Figure 2. Then, the force t using the triangular window is not limited to this, and the segment length T
It is preferable to set x to about 28m5ec.\The audio to be played is written after address IB in the buffer memory 22, and by playing this data with the DA converter 12, audio with varying speed is obtained. As described above, according to this embodiment, by providing the buffer memory 22 and the weighting calculation device 23, the reproduced sound is added multiple times, and the speed conversion is performed without missing data and eliminating the influence of discontinuity. In the explanation of this embodiment, a triangular window-like function is used as the weighting function.The quality of the reproduced sound will be improved if a Hamming window or a Hamming window-like weighting function is used. Also, the segment length Tx may be changed depending on the human input signal source.In general, for music sources, a longer segment length than the audio signal will produce better reproduced sound. According to the device, by providing a buffer memory and weighted addition calculation means, it is possible to read out and add data multiple times, and there is no data loss in special parts such as consonants, and natural speed conversion playback sound with no discontinuities. Also, according to the audio playback speed conversion method of the present invention, since pitch extraction is not used, relatively good reproduced sound can be obtained even for music in which the pitch period is difficult to obtain. This makes it possible to obtain an audio recording and reproducing device that is effective for rapid listening to audio, dictation of audio, etc.

[Brief explanation of drawings]

Fig. 1 shows the configuration of an audio recording/reproducing device according to an embodiment of the present invention. Fig. 2 (a) to (e) shows a schematic diagram of the speed conversion algorithm in the same embodiment. Fig. 3 shows the speed conversion ratio α. The flow of the data control procedure when ≦0.5 is as follows.
The figure shows the flow of the data control procedure in the case of speed conversion ratio α force (0°5 < α × 1).
Figure 6 shows the flow of the data control procedure when the speed conversion ratio α is 2≦α. Figure 7 shows the configuration of a conventional audio recording and reproducing device. Figure 8 shows the conventional configuration. FIG. 2 is a schematic diagram of a speed conversion algorithm. 11...AD conversion device 12...DA conversion device 1
3...Memory, 14...Data writing device 21.
...Data transfer device a 22...Buffer memory, 2
3... Weighted addition calculation device 24... Data reading device 25 Speed control device

Claims (5)

[Claims] (1) An AD conversion means, a memory for recording data, a data writing means for recording the AD-converted acoustic signal data in the memory, a buffer memory for temporarily storing the data, and the above-mentioned a data transfer means for transferring data from the memory to the buffer memory; and a data transfer means for reading the data on the buffer memory, and two
weighted addition calculation means for adding weighted data to a buffer memory and writing the result to a buffer memory; data reading means for reading data from the buffer memory; and DA conversion means for converting the output of the data reading means into an analog signal;
An audio recording and reproducing apparatus comprising: the data transfer means; the weighted addition calculation means; and a speed control means for controlling reading by the data read means according to a set reproduction speed. (2) An audio playback speed conversion method when the speed conversion ratio α defined by the following formula is 0<α≦05, the playback time length after speed conversionSpeed conversion ratio:α=playback time at normal speed The input signal to be subjected to long speed conversion processing uses time series data sampled in the time domain in advance, the predetermined fixed number of data used for speed conversion is T, and the output speed per one process is Assuming that the number of converted output data is T_0 given by the following formula, ▲ There are mathematical formulas, chemical formulas, tables, etc. ▼ Perform playback speed conversion processing using the following procedures (a) to (e). (a) From input signal data Read 2T■ data, (b) weight the first Y■ data with a monotonically decreasing function, weight the next T■ data with a monotonically increasing function, and calculate these two weighted data. (c) Of the weighted and added Y data,
Output the first Y_0 data as speed conversion output, (d) If there is not enough unprocessed input signal data to perform the same process again, terminate the process, otherwise perform the following process, (e) Of the T_■ data weighted and added, the remaining (T_■-
An audio playback speed conversion method characterized in that (T_■+T_0) pieces of unprocessed input signal data are added and read after T_0) pieces of data, and processing from (b) onwards is performed. (3) Speed conversion ratio α defined by the following formula is 0.5<α<1
An audio playback speed conversion method in which the speed conversion ratio is:
α = Playback time length after speed change / Playback time at normal speed The input signal to be processed for speed conversion uses time series data sampled in the time domain in advance, and a predetermined fixed value used for speed conversion is used. Let the number of data be T_■, and T_
Let x be the value given by the following formula, and let the number of speed conversion output data output per one process be (T_■ + T_x), T_x = 2α-1/1-α・T_■ The following (a) ~ Perform playback speed conversion processing according to procedure (e). (a) Read 2T_■ data from the input signal data, (b) Weight the first T_■ data with a monotonically decreasing function, and then T_■ pieces of data are weighted with a monotonically increasing function, these two weighted data are added to obtain T_■ pieces of data, and (c) the weighted and added T_■ pieces of data are subjected to speed conversion. (d) Add T_x pieces of unprocessed input signal data and read it and output it as speed conversion output; (e) When there is not enough unprocessed input signal data to perform the same process again. ends the process, otherwise (a
) An audio playback speed conversion method characterized by performing the following processing. (4) An audio playback speed conversion method when the speed conversion ratio α defined by the following equation is 1<α<2, where the speed conversion ratio: α=
Playback time length after speed conversion/Playback time at normal speed The input signal to be speed converted uses time series data sampled in the time domain in advance, and a predetermined fixed number of data is used for speed conversion. is T_■, T_x is the value given by the following formula, and the number of speed conversion output data output per one process is (2T_■ + T_x), T_x = 2-α/α-1・T_■ or less Perform playback speed conversion processing according to procedures (a) to (f) of (a) Read 2T_■ data from the input signal data, (b) First, directly use the first T_■ data as speed conversion output. (c) Next, the first T_■ data are weighted with a monotonically decreasing function, the next T_■ data are weighted with a monotonically increasing function, and these two weighted data are Add the data to obtain T_■ data and output it as a speed conversion output, (d) Then output the subsequent T_■ data as it is as a speed conversion output, (e) convert the unprocessed input signal data to (T_x -T_■) are added and output as speed conversion output, (f) If there is not enough unprocessed input signal data to perform the same process again, the process is terminated, otherwise (a
) An audio playback speed conversion method characterized by performing the following processing. (5) An audio playback speed conversion method when the speed conversion ratio α defined by the following formula is 2≦α, wherein the speed conversion ratio: α=: Playback time length after speed conversion/Playback time at normal speed The input signal to be subjected to long speed conversion processing uses time series data sampled in the time domain in advance, the predetermined fixed number of data used for speed conversion is T_■, and the speed output per one process is The number of data of conversion output is (T_■+
T_0) is given by the following formula, T_■=T_■/
α-1 Perform playback speed conversion processing using the following procedures (a) to (f): (a) Read 2T_■ data from the input signal data; (b) First, use the first T_■ data as is. (c) Next, the first T_■ pieces of data are weighted with a monotonically decreasing function, the next T_■ pieces of data are weighted with a monotonically increasing weighting function, and these two Adding the weighted data to obtain T_■ data; (d) Outputting the first T_■ data among the weighted and added T_■ data as a speed conversion output; (d) If there is not enough unprocessed input signal data to perform the same process again, the process is terminated; otherwise, the following process is performed; After the remaining (T_■-T_■) pieces of data that were not present, the latter half of the T_■ pieces of data read in (a) are placed. (f) T_■ pieces of unprocessed input signal data are added and read, (b) A speed conversion method characterized by performing the following processing.