JPH03198000A - Voice time-base compressing and encoding device - Google Patents

Abstract

PURPOSE:To prevent a voice from deteriorating by using a compressed waveform signal as a voice waveform signal with a constant pitch cycle and connecting the remaining voice waveform signal to the compressed waveform signal without compression. CONSTITUTION:An in-frame time-base compressing means 4 sections an input voice waveform 3 into constant-length analytic frames, and pitch cycles in the analytic frames are found by, for example, autocorrelation analysis and utilized to compress voice waveforms in the analytic frames on the time basis, thereby finding a compressed waveform 5. Further, this device is equipped with an in-frame time-base expanding means 15 which expands the compressed waveform 14 into a voice waveform with original analytic frame length by utilizing the pitch cycles 6 after the signal is decoded in analytic frame units. Then a frame which is longer than plural pitches of the input voice waveform is set, a waveform of plural (p) pitches in this frame is compressed into a waveform of less integral (q) pitches, and the remaining voice waveform signal is connected to the compressed waveform with the integral pitch without being compressed. Consequently, even if a transmission error of pitch cycles is generated, its influence stays in the sectio.

Description

[Detailed Description of the Invention] [Industrial Application Field] The present invention relates to an audio time axis compression/encoding device that performs time axis compression/expansion of an audio waveform based on a pitch period within a fixed length analysis frame range. .

[Conventional technology]

A conventional device of this type is shown in FIG. The reason for this is I OC'82 (IEEE Int
national communication
Paper presented at C1onference, 19B2) R, V, 0oxeta'1. 'An Impleme
ntation of Time Domain Har
monic scaling with application
ation to 5peech Coding” (1
0082, pp. 4G, 1.1-4). In the figure, (1) is a microphone, (2) is an A/D converter, and (3) is a Input audio waveform, Q Samurai is time axis compression means, (E) is pitch period extraction means, (5) is compressed waveform, (6) is pitch period, (7)
is an encoding means, (8) is a pitch encoding means, (9) is a multiplexer, Ql is a transmission line, (b) is a demultiplexer, (2) is a decoding means, (2) is a pitch decoding means,
Q4 is a compressed waveform after decoding, (F) is a time axis expansion means, α is an expanded waveform, aη is a D/A converter, and (To) is a speaker.

[Operation of conventional device]

Next, the operation will be explained.First, the transmitter shown in FIG. 4 will be explained. Voice input from the microphone (1) is converted into a discrete data series (hereinafter referred to as input voice waveform (3)) by an A/D converter (2). The pitch period extraction means obtains the pitch period (6) of the input speech waveform (3), for example, by autocorrelation analysis. The time axis compression means compresses the input speech waveform (3) to 1/2 on the time axis for every two consecutive pitch sections based on the pitch period (6) extracted by the pitch period extraction means Kan, and generates a compressed waveform. Find (5). Here, the details of the time axis compression performed by the time axis compression means α scratch are explained in the fifth section.
In Figure 5, (5a) 4-i consecutive 2
Input audio waveform of pitch section 0 (n) (Ob)
is a compressed waveform 5C(n) obtained by compressing the input audio waveform S (n) by 1/2 on the time axis. Now input audio waveform 8 (n)
If the pitch period of is Tp, the compressed waveform 5c(n) is the following peak, 03c(n)=(1-
Wc(n) )-5(n)+ Wc(n) ・S(n
+ Tp ) tl) wc(n) = (n −1/
2)/Tp t2)n=1,2
．． *zjT.

Here, (II), WC(n) in equation (2) is a window function that is applied to the input audio waveform 5(n), and in Figure 5, the input audio waveform S (n) is multiplied by the window function wc(n) and compressed. Waveform 5c (
The procedure for finding n) is shown.

Now, the compressed waveform (5) obtained by the above method using the time axis compression means a value shown in FIG. It is encoded by the encoding means (8). The multiplexer (9) multiplexes the encoded compressed waveform signal and the pitch period signal and outputs the multiplexed signal to the transmission line (In).

Next, the receiving section shown in FIG. 4 will be explained. The encoded compressed waveform i*=i and the pitch period information propagated through the transmission path Ql are separated by a demultiplexer t)1).

The compressed waveform information is decoded by the decoding means (2) in a decoding process corresponding to the encoding means (7), and becomes a compressed waveform α→ after decoding. Similarly, the pitch period information is also decoded by the pitch decoding means 0. The time axis expansion means 121) expands the pitch period decoded by the pitch decoding means qg ((based on il, decoded 18E contraction C-4 in pitch period units) on the time axis, Determine the expanded waveform OQ. The total time axis expansion performed by the time axis expansion means 12]J, !tI, is shown in FIG. 6. In FIG. r, r+1.th) decoded compressed waveform sum c
(n), and (6b) is the compressed waveform after decoding ◇ An expanded waveform ◇ in which the r-th pitch section of c(n) is expanded by twice. (n
). Now, if the starting point of the r-th pitch section in the decoded compressed waveform c(n) is set to n=1, the expanded waveforms e and (n) are determined by the following equations (3) and (4).

G, (n) = <1- We(n) > Cypress c(n)+
we(n) 4c(n −Tp> (3) We (
n)=(n-1/2)/2T p
(4) n=1,2. ...2Tp (3), we(n) in equations (4) is the royal capital waveform after decoding◇
It is a window function that is applied to c(n), and in FIG.
←).

The expanded waveform αQ obtained by the above method by the time axis expansion means 31) in FIG. [Problem to be Solved by the Invention] The conventional audio time axis compression encoding device is configured as described above, and the time axis expansion means of the decoding section is used for decoding. 1kE
Since the E-condensed waveform is sequentially expanded to twice the length of the pitch period transmitted from the transmission line, if one pitch period is erroneously transmitted due to a transmission error that occurs on the transmission line, it is limited to only that pitch period section. There is a problem in that the error spreads to all expanded waveforms that are subsequently expanded by the time axis expansion means, resulting in significant deterioration in sound quality.

Furthermore, when the audio encoding/decoding process performed by the encoding/decoding means is performed in units of analysis frames of a certain length, the analysis frame interval and the interval of time axis compression/expansion performed in pitch cycle units cannot be matched. Therefore, a problem arises in that the audio encoding/decoding process and the compression/decompression process must be performed separately in separate sections, making the process complicated and potentially causing a processing delay.

[Means to solve the problem]

The audio time @compression encoding device according to the present invention includes a time axis compression means for compressing the time axis of an audio waveform based on a pitch period within the range of the analysis frame for each analysis frame of a fixed length; The apparatus is equipped with a time axis expansion means for expanding the time axis of the time axis compressed waveform to the original analysis frame length.

[Effect]

The time axis compression means according to the present invention, for each analysis frame of a certain length, converts the audio waveform of several pitch periods continuously existing within the analysis frame into two pitch periods in every two non-contiguous pitch period sections. Compression is performed on the time axis using the similarity of the waveforms of the sections.

Further, the time axis expansion means in the present invention expands the compressed audio waveform, which has been time axis compressed for each analysis frame by the time axis compression means, on the time axis based on the pitch circumference side to form the audio waveform of the original analysis frame length. Make it.

〔Example〕

FIG. 1 is a block diagram showing an embodiment of the present invention. 1
1-(31, (5)-04), M-J119 is the 4th
In FIG. 1, (4) is an intra-frame time axis compression means, and (g) is an intra-frame time axis expansion means, which is the same as the conventional apparatus shown in the figure, and the explanation thereof will be omitted.

In Fig. 1, the intra-frame time axis compression means (4) divides the input speech waveform (3) into analysis frames of a certain length, determines the pitch period within each analysis frame by, for example, autocorrelation analysis, and utilizes this pitch period. Then, the audio waveform within the analysis frame is compressed on the time axis to obtain a compressed waveform (5).

Furthermore, the aggressive pitch period (6) is determined by the pitch encoding means (8).
) is encoded.

For each analysis frame, the intra-frame time axis expansion means expands the decoded compressed waveform 0.0 into an audio waveform having the original analysis frame length using the pitch period (6).

Figures 2 and 3 show the intra-frame time axis E in this embodiment.
The details of the time axis compression/expansion for each analysis frame executed by the EM means (4) and the intra-frame time axis expansion means (to) are shown.

Figure 2 shows time axis compression/expansion when the pitch period T of the input speech waveform is N/4 (Tp≦N/3) with respect to the analysis frame length N. As a condition for performing late-night axis EE reduction, it is essential that the analysis frame length N is at least twice the maximum pitch period of 1°, (max).

The intra-frame time axis compression + stage (4) uses the input audio waveform (2
The first 2-pitch period section of a) is +11 above,
t2) The remaining audio waveform of the 0 analysis frame whose time is compressed to 1 pitch period by the formula does not have the length of 2 pitch period sections, so time axis compression of this section is performed.Time axis compressed audio waveform The FE shrunken skin shape (2b) is obtained by piecing together the two as they are. Here, the input audio waveform is (N − T,
)/'H.

The intra-frame time axis expansion means αe generates the decoded FEII waveform (
The audio waveform for the first 1 pitch period of 2C) is
31, (4) Expand it to 2 pitch periods by
Find the expanded waveform (2(1)) with the original analysis frame length N.

FIG. 3 shows the cycle time@compression/expansion when the pitch period Tp of the input speech waveform satisfies N15 <Tp <- with respect to the analysis frame length N.

In this case, the intra-frame time axis compression means (4) converts the first and second 2-pitch period sections of the input audio waveform (3a) into the above +1.1, !, respectively. 2) The time axis is compressed to one pitch period using the formula. As in the case of FIG. 2, the remaining audio waveforms of the analysis frame are not subjected to time axis compression, but are simply spliced to the time axis compressed audio waveform to obtain a compressed waveform (3b). Here, the input audio waveform is (N −2
Tp)/N K compressed.

The intra-frame time axis expansion means first decodes the compressed waveform (
3C) first pitch period section (■ in Figure (3c))
is expanded twice using equations (3) and (4) above.

Next, the second pitch period section (■ in Figure (3C)) is
A replacement waveform (3d) in which the non-existent third pitch period section is replaced with the second pitch period section is used to expand the waveform twice. Then, the decoded audio waveform (3
The doubled expanded waveform of the first and second pitch period sections in C) and the remaining section of the decoded audio waveform (3C) are connected to obtain the expanded waveform (3e). At the beginning, as shown in FIG. 3, time axis FE and compression/expansion are performed using the pitch period within an analysis frame of a constant length.

Next, since the time axis compression rate of the in-frame time axis compression means according to the present invention changes depending on the pitch period, let us consider the change in this time axis compression rate. Let the number be k, the analysis frame length N, and the pitch period T. k is the number of times of time axis compression performed in units of two pitch period sections within the analysis frame, and TpXk is the length of time axis compression actually performed. Therefore, the time axis compression ratio R is as follows.

The range in which the pitch period T and the time axis compression rate exist when k=1.2.3 is shown below. From equation (7), the range in which the time axis compression 4R exists can be formulated as the following equation.

And the time axis compression ratio R at that time is. Therefore, the intra-frame time axis compression means according to the present invention can time axis compress the input audio waveform within the analysis frame within the range of the time axis compression ratio shown by equation (9).

In addition, in the above embodiment, the encoding means (power) performs pulse encoding, but any other audio encoding method may be used. When used, the analysis frame set by the intra-frame time axis compression means and the analysis frame of this encoding method may be the same.In this case, the decoding method in the decoding means α follows the encoding method. [Effects of the Invention] As described above, according to the present invention, since the time axis compression/expansion of the audio waveform based on the pitch period can be performed within the analysis frame range of a certain length, transmission errors on the transmission path can be prevented. Even if the pitch period of a certain analysis frame is erroneously transmitted, the direct effect of this error on the expanded waveform can be suppressed within that analysis frame, and moreover, the encoding means performs voice encoding for each analysis frame of a certain length. When this is done, this analysis frame and the analysis frame of the time axis compression/expansion are made the same, thereby simplifying the processing.

[Brief explanation of drawings]

FIG. 1 is an overall configuration diagram showing an embodiment of the audio time axis compression encoding device of the present invention, FIGS. 2 and 3 are waveform explanatory diagrams showing processing in an embodiment of the invention, and FIG. 4 is an overall configuration diagram showing an example of a conventional audio time axis compression encoding device, FIG. 5 is an explanatory diagram showing the time axis compression process of the conventional audio time axis compression encoding device, and FIG. 0 is an explanatory diagram showing the time axis expansion process of the compression encoding device. In the figure, (1) is the microphone, (2) is the A/D conversion i, (3) is the input audio waveform, and (4) is the time within the frame. Axial compression means, (5) compression waveform, (6) pitch period,
(7) is an encoding means, (8) is a pitch encoding means, (9
) is the multiplexer, Ql is the transmission line, (11) is the demultiplexer, (a) is the decoding means, (to) is the pitch decoding means, α→ is the compressed waveform after decoding, (to) is the time axis within the frame Decompression means, a. is the decompressed waveform, Q'h is the D/A converter, (to) is the speaker, (2a) is the input audio waveform, (2b
) is the compressed waveform, (20) is the compressed waveform after decoding, (2d) is the expanded waveform, (3a) is the input audio waveform, (3b) is the compressed waveform, (3C) is the compressed waveform after decoding, (3d) is the replacement waveform,
(3e) is an expanded waveform, α9 is a time axis compression means, (1) is a pitch period extraction means, @E is a time axis expansion means, (5a)
is an input audio waveform, (5b) is a compressed waveform, (6a) is a compressed waveform after decoding, and (61) is an expanded waveform. Note that the same reference numerals in each figure indicate the same or corresponding parts.

Claims (1)

[Claims] In an audio time-base compression/expansion device that consists of a transmitter that compresses and encodes an input audio waveform on the time axis, and a receiver that decodes and expands the compressed-encoded audio waveform on the time axis, the input audio The transmitting unit includes an intra-frame time axis compression means for compressing the waveform on the time axis within the range of the analysis frame for each analysis frame of a fixed length by using the pitch period in this analysis frame of the input audio waveform, The reception section is characterized by comprising an intra-frame time axis expansion means for expanding the audio waveform time-axis compressed by the time axis compression means into an audio waveform having the analysis frame length based on the pitch period for each analysis frame. Audio time axis compression encoding device.