JP2004266587A - Time-sequential signal encoding apparatus and recording medium - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a time-sequential signal encoding apparatus and recording medium in which digitally edited high-definition audio work data can be efficiently reversibly compressed. <P>SOLUTION: Time-sequential sample streams are divided into odd-numbered and even-numbered streams, a differential value between an odd-numbered sample and an average of both neighboring even-numbered samples and a differential value between an even-numbered sample and an average of both neighboring odd-numbered samples are calculated, the stream with continuous smaller differential values is relocated as a main sample stream and the other stream is relocated as a sub sample stream. With respect to the relocated main sample stream and sub sample stream, a predictive error value based on a linear predictive error is obtained, and variable length encoding is performed. <P>COPYRIGHT: (C)2004,JPO&NCIPI

Description

[0001]
[Industrial applications]
INDUSTRIAL APPLICABILITY The present invention relates to music production fields such as music production, storage of acoustic data materials, relay of location materials, audio recording / reproduction using digital recording media such as CD / DVD, and analysis / diagnosis of biological signals in telemedicine. The present invention relates to a suitable data compression encoding technique.
[0002]
[Prior art]
Conventionally, various methods have been used for compressing an acoustic signal. MP3 (MPEG-1 / Layer3), AAC (MPEG-2 / Layer3) and the like have been put to practical use as a technique for compressing and encoding an audio signal. With such a compression encoding method, it is possible to treat an audio signal as small data, which contributes to the efficiency of data recording and transmission.
[0003]
Recently, not only lossy coding schemes such as MP3 and AAC as described above, but also lossless coding schemes that can be completely restored have been developed, and are used for sound material management (for example, Patent Document 1).
[0004]
[Patent Document 1]
JP 2000-821199 A
[0005]
[Problems to be solved by the invention]
When handling high-definition audio, sampling is performed with the sampling frequency or the number of quantization bits set higher than that of normal audio (music CD quality: sampling frequency 44.1 kHz, quantization bit number 16 bits). In the music editing field, the task of mixing and editing various sounds is often performed. In this case, normal audio may be partially mixed, and this normal audio is adjusted to high definition audio To this end, interpolation of samples, extension of the number of quantization bits, and the like are performed. The sound signal that has been subjected to such interpolation processing can be compressed in principle to the information amount of the original sound. However, even if encoding is performed by the conventional lossless coding method, the compression is not so large. No compression effect is obtained. Conversely, if compression is performed directly with linear prediction coding as it is, the linear prediction error at the interpolated location may increase and the compression ratio may deteriorate, which may have the opposite effect of compressing the redundant location. Can be. For example, audio data whose sampling frequency has been doubled can theoretically be compressed to 50% or less, but in reality, the compression ratio exceeds 50%.
[0006]
Therefore, in order to solve such a problem, the present invention provides an encoding device and a recording medium for a time-series signal capable of efficiently reversibly compressing high-definition audio work data in the course of digital editing. As an issue.
[0007]
[Means for Solving the Problems]
In order to solve the above-mentioned problem, in the present invention, for a time-series signal composed of a time-series sample sequence, an encoding device that compresses the amount of information so that all the sample sequences can be reproduced, In contrast, a main sample sequence created by recording and a sub-sample sequence obtained by interpolating the main sample sequence are separated, and the value of each sub-sample in the sub-sample sequence is replaced with a neighboring main sample sequence. A sample sequence rearrangement means for converting the average value of the sample and the value of the subsample into a difference value, calculating a linear prediction error for each of the main sample sequence and the converted subsample sequence, Prediction error conversion means for converting the values of the sample sequence and the sub-sample sequence into prediction error values, respectively, and encoding the main sample sequence and the sub-sample sequence converted into the prediction error value with a variable length. Characterized by being configured to have a variable length coding means.
[0008]
According to the present invention, the sample sequence constituting the time-series signal is rearranged into a main sample sequence obtained by recording and a sub-sample sequence obtained by interpolating the main sample sequence, and the main sample sequence and the sub-sample sequence are rearranged. Since each sample sequence is predictively coded and variable-length coded, work data of high-definition audio in the process of digital editing can be efficiently and reversibly compressed.
[0009]
BEST MODE FOR CARRYING OUT THE INVENTION
Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.
(1st Embodiment)
FIG. 1 is a configuration diagram showing a first embodiment of a time-series signal encoding apparatus according to the present invention. In FIG. 1, reference numeral 10 denotes a time-series signal input unit, 20 denotes a sample sequence rearrangement unit, 30 denotes a lower fixed bit deletion unit, 40 denotes a prediction error conversion unit, 50 denotes an inter-channel calculation unit, 60 denotes a polarity processing unit, and 70 denotes It is a variable length encoding means.
[0010]
In FIG. 1, a time-series signal input unit 10 has a function of inputting a digitized sound signal such as a digital sound signal. The sample sequence rearranging means 20 converts a sample sequence, which is an input time-series signal, into a main sample sequence, which is a sample sequence obtained based on the recording, and a sub-sample sequence, which is obtained by interpolating the main sample sequence. It has the function of separating. The lower fixed bit deletion unit 30 has a function of deleting a predetermined number of lower bits that are quantization noise components. The prediction error conversion means 40 has a function of converting the value of each sample into a prediction error value using a linear prediction error technique. The inter-channel calculation means 50 has a function of performing a difference calculation between channels of a sample sequence including a plurality of channels. The polarity processing unit 60 has a function of dividing a bit string of each sample, which represents a positive / negative value in a complement expression, into one bit representing positive / negative polarity and another bit string. The variable length coding means 70 has a function of coding the value of each sample with a variable bit length. The apparatus shown in FIG. 1 is actually realized by a computer and a dedicated software program installed on the computer.
[0011]
Next, the processing operation of the time-series signal encoding device shown in FIG. 1 will be described. In the present invention, a case will be described as an example in which work data in which a plurality of audio signals are mixed is handled as a time-series signal. As described above, such an audio signal includes both a normal audio signal having a sampling frequency of 48 kHz and a quantization bit number of 16 bits and a high-definition audio signal having a sampling frequency of 96 kHz and a quantization bit number of 24 bits. . A sound signal obtained by mixing sound signals having different sampling frequencies in this manner is handled by unifying the sampling frequency to a high-definition sound signal. In this case, a sound signal having a sampling frequency of 48 kHz is interpolated between the sound signals having a sampling frequency of 96 kHz with an average value of adjacent samples in order to match the number of samples.
[0012]
FIG. 2A schematically shows such an acoustic signal. In FIG. 2A, numbers in parentheses are sample numbers assigned in ascending order from 1, and x indicates a value of the sample. When such a time-series signal as a sample sequence is input from the time-series signal input means 10, the sample sequence rearrangement means 20 calculates the difference between the odd-numbered sample and the average value of the adjacent even-numbered samples. And the difference between the even-numbered sample and the average value of the adjacent odd-numbered sample is calculated. The sample rows at this time are schematically shown in FIGS. 2B and 2C, respectively. In the example of FIG. 2B, the even-numbered samples that are not subjected to the calculation are shown in the state of being moved temporally in the past, and in the example of FIG. .
[0013]
As a result of this difference operation, the one with the larger difference value is set as the sub-sample sequence, and the one with the smaller difference value is set as the main sample sequence. In the example of FIG. 2, each value of the latter half of the array of FIG.
The values in the second half of the array in FIG. 2C are compared. For example, when the odd-numbered sample is obtained by interpolation, the latter half value of the array shown in FIG. On the other hand, when the even-numbered samples are obtained by interpolation, the latter half value of the array shown in FIG. For example, when there are many values near 0 in the latter half of the array shown in FIG. 2B, the set of even-numbered samples is separated as the main sample sequence, and the set of odd-numbered samples is separated as the sub-sample sequence. In the processing of the sample sequence rearrangement means 20, the main sample is moved in the past in time and the sub-sample is moved in the future in time, as shown in FIGS. Alternatively, the main sample and the sub-sample may be handled separately. For example, if the odd number is a sub-sample, the main sample and the sub-sample are separated as shown in FIG. In the present invention, by performing linear prediction on a sample sequence including samples obtained by interpolating using original samples, the main sample Columns and subsample columns. Therefore, if linear prediction can be separately performed on the main sample sequence and the sub-sample sequence, even if only one sample sequence as shown in FIGS. May be two sample strings as shown in FIG.
[0014]
Next, the lower fixed bit deletion unit 30 separates a predetermined number of lower bits of each sample of the main sample sequence and the sub-sample sequence. This is performed in order to remove redundant lower-order bit components when the data having a quantization bit number of 16 bits is converted into 24 bits in order to match with a high-definition audio signal. , The amount of coded information will increase by a factor of 3/2. On the other hand, if all the audio signals of the source material to be mixed are quantized with high-definition 24 bits, it is not necessary to perform the processing by the lower fixed bit deleting means 30. A means for separately recording the obtained lower bit data array as a part of the output code data reduces the processing load after the subsequent prediction error conversion means. Regarding the lower fixed bit deleting means 30, whether or not to operate can be set in advance.
[0015]
Subsequently, the prediction error conversion means 40 converts the values of each sample of the main sample sequence and the sub-sample sequence into prediction error values. The calculation of the prediction error value for a certain sample is performed using the value of one or a plurality of samples located immediately before in the past in time. In the present embodiment, a method of dynamically changing the number of samples immediately before use is used. Hereinafter, such adaptive linear prediction encoding will be described. FIG. 3 is a flowchart showing an outline of processing of adaptive linear prediction encoding performed by the prediction error conversion means 40. First, using a plurality of prediction formulas prepared in advance, a linear prediction error corresponding to each prediction formula is calculated (step S1). Specifically, the following [Equation 1] to [Equation 6] are prepared as prediction calculation equations for calculating the prediction error of the sample number t.
[0016]
[Formula 1]
e0 (t) = x (t) -e0 (t-1) / 2
[0017]
[Formula 2]
e1 (t) = x (t) -a₁₁X (t-1) -e1 (t-1) / 2
[0018]
[Equation 3]
e2 (t) = x (t) -a₂₁X (t-1) -a₂₂X (t-2) -e2 (t-1) / 2
[0019]
[Equation 4]
e3 (t) = x (t) -a₃₁X (t-1) -a₃₂X (t-2) -a₃₃X (t-3) -e3 (t-1) / 2
[0020]
[Equation 5]
e4 (t) = x (t) -a₄₁X (t-1) -a₄₂X (t-2) -a₄₃X (t-3) -a₄₄X (t-4) -e4 (t-1) / 2
[0021]
[Equation 6]
e5 (t) = x (t) -a₅₁X (t-1) -a₅₂X (t-2) -a₅₃X (t-3) -a₅₄X (t-4) -a₅₅X (t-5) -e5 (t-1) / 2
[0022]
In the above [Equation 1] to [Equation 6], e0 (t) to e5 (t) are prediction errors in the sample at time t by the respective prediction calculation formulas, and x (t) to x (t-5) are time Sample values at t to t-5.
[0023]
[A]₂₁X (t-1) + a₂₂X (t−2) ”and“ a ”in the above [Equation 4].₃₁X (t-1) + a₃₂X (t-2) + a₃₃X (t−3) ”and“ a ”in the above [Equation 5].₄₁X (t-1) + a₄₂X (t-2) + a₄₃X (t-3) + a₄₄X (t−4) ”and“ a ”in the above [Equation 6].₅₁X (t-1) + a₅₂X (t-2) + a₅₃X (t-3) + a₅₄X (t-4) + a₅₅"X (t-5)" is a linear prediction component based on the past 2 to 5 samples. The prediction error at time t using this linear prediction component and the prediction errors “e1 (t−1) / 2” to “e5 (t−1) / 2” (error feedback components) calculated in the immediately preceding sample. e0 (t) to e5 (t) are calculated.
[0024]
The above coefficient a₁₁~ A₅₅Has an initial value of a₁₁= 1, a₂₁= 2, a₂₂= -1, a₃₁= 3, a₃₂= -3, a₃₃= 1, a₄₁= 4, a₄₂= -6, a₄₃= 4, a₄₄= -1, a₅₁= 5, a₅₂= -10, a₅₃= 10, a₅₄= -5, a₅₅= 1 are set, but in the present embodiment, these coefficients are dynamically changed. Specifically, the coefficient a is calculated using the following [Equation 7] using the Levinson-Durvin algorithm.₁₁~ A₅₅To determine.
[0025]
[Equation 7]
φ (k) = 1 / (NK) −_{j = 1, NK}x (j) · x (j + k)
k_i= − {Φ (i) +}_{j = 1, i-1}a_j(I-1) · φ (ij)｝ / E (i-1)
a_i(I) = k_i
a_j(I) = a_j(I-1) + k_i・ A_ij(I-1) where 1≤j≤i-1
E (i) = (1-k_i ²) E (i-1)
[0026]
In the above [Equation 7], φ (k) is a sample obtained by shifting k samples within a range of the maximum value K (5 in the above example) in N samples x (j) (j = 1,..., N). The autocorrelation value with the column. Note that N takes a sufficiently large value with respect to K (for example, when K = 5, N = 32768). [Equation 7] is recursively repeated from i = 1 to i = K, and finally obtained a_j(K) becomes the coefficient corresponding to the past K samples, and a is an intermediate result obtained in each phase._j(I) is the coefficient a_ijBecomes In step S1, a calculation is performed using each of the formulas [1] to [6] using the coefficient determined by the above [7]. The calculation based on [Equation 7] is actually performed in step S7 described below. Further, since the values of several past samples are required to determine the coefficients, the calculation of [Equation 1] to [Equation 6] is performed for the first N-1 samples using the above initial coefficients. become.
[0027]
Next, a linear prediction error that minimizes the cumulative error, which is the accumulation of the absolute values of the prediction error values for each of the above-described prediction formulas, is selected as the prediction error of the sample (step S2). Here, the concept of accumulated error is used. Specifically, the accumulated values of the prediction errors calculated by the prediction calculation formulas [Formula 1] to [Formula 6] for the past samples are set as A0 to A5. Then, a prediction error corresponding to the smallest one of the accumulated errors A0 to A5 is selected. For example, it is assumed that A2 is the smallest among A0 to A5. In this case, the prediction error e2 (t) calculated by [Equation 3] is selected as the prediction error e (t) to be encoded. The selected prediction error e (t) is replaced with the original value x (t) of the sample, and the subsequent processing is performed.
[0028]
Subsequently, the absolute values of the prediction errors e0 (t) to e5 (t) are added to the accumulated errors A0 to A5 (step S3). Specifically, as shown in the following [Equation 8], variables A0 to A5 serving as accumulated error values are updated. At the same time, every time the processing of each sample is performed, the processing of adding the counters C1 and C2 one by one is performed.
[0029]
[Equation 8]
A0 ← A0 + | e0 (t) |
A1 ← A1 + | e1 (t) |
A2 ← A2 + | e2 (t) |
A3 ← A3 + | e3 (t) |
A4 ← A4 + | e4 (t) |
A5 ← A5 + | e5 (t) |
[0030]
Subsequently, it is determined whether or not the counter C1 has exceeded a predetermined number (step S4). In the present embodiment, the predetermined number is set to 100 times. That is, it is determined whether the counter C1 has exceeded 100.
[0031]
As a result, if the counter exceeds 100, the accumulated error is halved (step S5). Specifically, as shown in the following [Equation 9], variables A0 to A5 that are cumulative errors are divided by two. At the same time, the counter C1 is reset to 0. That is, A0 to A5 here are not accumulated errors in a pure sense, but are moving averages of the accumulated errors. In the present embodiment, up to the immediately preceding maximum of 100 samples are accumulated, but the previous samples are processed so as to be halved. Thereby, the influence of the samples separated in time is reduced.
[0032]
[Equation 9]
A0 ← (A0) / 2
A1 ← (A1) / 2
A2 ← (A2) / 2
A3 ← (A3) / 2
A4 ← (A4) / 2
A5 ← (A5) / 2
[0033]
Subsequently, it is determined whether or not the counter C2 has exceeded a predetermined number (step S6). In the present embodiment, the predetermined number is set to 32768 times. That is, it is determined whether the counter C2 has exceeded 32768.
[0034]
As a result, if the counter C2 exceeds 32768, the coefficient a₁₁~ A₅₅Is calculated again (step S7). Specifically, using the above [Equation 7], the coefficient a₁₁~ A₅₅Will be recalculated. At the same time, the counter C2 is reset to zero.
[0035]
By executing the processes of steps S1 to S7 over the samples of the main sample sequence and the sub-sample sequence, the values of all the samples are replaced from the original amplitude value x (t) by the target error e (t). become. In the present embodiment, in particular, it is possible to calculate a more accurate prediction error by dynamically changing coefficients of a plurality of prediction formulas.
[0036]
Next, the inter-channel calculation means 50 performs a difference calculation between the channels on the main sample and the sub-sample of each channel in which the prediction error value is recorded. FIG. 4 is a flowchart showing the outline of the process of calculating the difference between channels. First, a main sample is read (step S11). If the read main sample is for the L channel, the value of the main sample is stored in the memory and output to the next polarity processing means 60 (step S12). On the other hand, if the read main sample is for the R channel, the processing from step S13 is performed. Note that the main sample is read alternately for each channel, so there is no need to make a determination. For example, once the sample of the L channel with the sample number t = 1 is read, the order of the sample of the R channel with the sample number t = 1 is determined next, and then the sample of the L channel with the sample number t = 2. Therefore, the processing in step S12 and the processing after step S13 may be alternately switched.
[0037]
In the case of the R channel sample, the variables Ao and Ad are compared (step S13). Here, Ao is the accumulation of the absolute values of the sample values of the R channel, and Ad is the accumulation of the absolute values of the differences between the sample values of the R channel and the L channel. The initial values of both variables Ao and Ad are 0. If Ad ≧ Ao in step S13, the sample of the R channel is output as it is to the next polarity processing means 60 (step S14). Further, the accumulated value Ao is updated as shown in the following first equation of [Equation 10]. Specifically, the sample value e of the R channel_RIs added to the accumulated value Ao.
[0038]
If Ad <Ao in step S13, the difference from the L channel sample stored in the memory in step S12 is calculated, and the difference value is output to the next polarity processing unit 60 (step S15). Further, the accumulated value Ad is updated as shown in the following Expression (10). Specifically, the difference value e between the samples of the R channel and the L channel_R-E_LIs added to the accumulated value Ad.
[0039]
[Equation 10]
Ao ← Ao + | e_R|
Ad ← Ad + | e_R-E_L|
[0040]
Subsequently, the counter C3 that indicates that a pair of L and R samples has been processed is incremented by one (step S16).
[0041]
Subsequently, it is determined whether the counter has exceeded a predetermined number of times (step S17). In the present embodiment, the predetermined number is set to 100 times. That is, it is determined whether or not the counter C3 has exceeded 100.
[0042]
As a result, if the counter C3 exceeds 100, the cumulative values Ao and Ad are halved (step S18). Specifically, as shown in the following [Equation 11], variables Ao and Ad, which are cumulative values, are divided by two. At the same time, the counter C is also reset to half. That is, the variables Ao and Ad here are not cumulative values in a pure sense, but are moving averages of the cumulative values, similarly to A0 to A5 in [Equation 8]. In the present embodiment, up to the immediately preceding maximum of 100 samples are accumulated, but the previous samples are processed so as to be halved. Thereby, the influence of the samples separated in time is reduced.
[0043]
[Equation 11]
Ao ← Ao / 2
Ad ← Ad / 2
C3 ← C3 / 2
[0044]
By executing the processing of steps S11 to S18 over all the main samples in the main sample sequence, the values of all the samples of the R channel are replaced with the difference values from the L channel. However, as is apparent from the above-described processing, depending on the magnitude relationship between the accumulated values Ao and Ad, there is a sample in which the sample value of the R channel is recorded as it is. In the inter-channel calculation means 50, all the samples of the L channel are output as they are. When the processing is completed for each sample in the main sample sequence, the same processing is performed for the sub-samples. The processing by the inter-channel calculation means 50 is omitted for a monaural sound signal having only one channel.
[0045]
Subsequently, the polarity processing means 60 performs positive / negative processing on each sample. By the processing of the prediction error conversion means 40 and the inter-channel calculation means 50, the value of each sample was replaced from the amplitude value to the prediction error, and the value of the R channel was replaced by the difference from the L channel. Is unchanged from the original. Normally, when operated by a computer such as a computer, each data is processed in units of 32 bits and is expressed using a two's complement representation. This is converted into a positive / negative signed absolute value expression, and the absolute value portion is shifted one bit higher, and the positive / negative sign bit is shifted to the LSB (least significant bit). FIG. 5 schematically shows the conversion of the bit configuration by the polarity processing means 60. FIG. 5A shows a bit configuration before processing, and FIG. 5B shows a bit configuration after processing. The reason why the positive / negative sign bit is shifted to the LSB in this way is to make it easier to detect the bit length of each sample in the subsequent processing of the variable length coding means 70.
[0046]
Next, the variable length coding means 70 performs a process of converting each sample into a variable length. The variable length coding in the present embodiment employs a method generally called Golomb coding. More specifically, the bit components forming one sample are divided into upper bit components and lower bit components, the lower bit components are left unchanged, and the upper bit components are the numerical values obtained by converting only the upper bits into decimal numbers. Bits “0” are arranged, and a separator bit “1” is added at the end. For example, consider an 8-bit bit component "00101000". At this time, if the lower bit component is 4 bits, the lower bit component is “1000”. Since the upper bits are "0010", "2" pieces of "0" obtained by converting this number into decimal numbers are arranged, and are converted into "001" obtained by adding "1" at the end. As a result, the 8-bit bit string “00101000” is converted into a 7-bit bit string “00111000”. In the present embodiment, the bit length of the lower bit component that makes the bit component invariable before and after the conversion is made variable in each sample.
[0047]
Hereinafter, the processing performed by the variable length coding unit 70 will be specifically described. FIG. 6 is a flowchart showing an outline of the variable length coding. First, an average bit length Bf which is a moving average of bit lengths of past samples is calculated (step S21). The average bit length Bf is obtained by dividing the cumulative bit length RB, which is the cumulative value of the past bit length, by a counter C4 based on the number of past samples. That is, it is calculated by Bf = RB / C4. Since the accumulated bit length RB is 0 in the initial state, when processing a sample at t = 1, the bit length Bd (t) of the sample at t = 1 is set as an initial value. Also, the initial counter C4 is set to 1.
[0048]
Subsequently, the bit length Bd (t) of the sample at time t is calculated (step S22). For the samples after t = 2, after calculating the average bit length Bf, the bit length Bd (t) of the sample is calculated. The bit length Bd (t) can be easily calculated by converting the bit configuration by the polarity processing means 60. As a result of the conversion into the bit configuration as shown in FIG. 5B, the bit length starts from the bit “1” appearing at the head in the bit configuration of each sample. Next, the bit length Bv of the changing unit is calculated (Step S23). This is calculated by subtracting the average bit length Bf from the bit length Bd (t) of the sample. Subsequently, data sign output is performed (step S24). Specifically, after outputting “0” by the numerical value obtained by converting the upper Bv bits into a decimal number, a separator bit “1” is output, and the lower Bf bits are output as an invariable part. The code output is performed as recording on an external storage device such as a hard disk or a CD-R. Next, the bit length Bd (t) is added to the accumulated bit length RB (step S25). At the same time, every time the processing of each sample is performed, the processing of adding the counter C4 one by one is performed. Subsequently, it is determined whether the counter C4 has exceeded a predetermined number (step S26). As the predetermined number, about 100 is set here as well. Therefore, it is determined whether the counter 4 has exceeded 100. As a result, if the counter exceeds 100, the cumulative bit length RB is halved (step S27). Specifically, the variable RB that is the accumulated bit length is divided by two. At the same time, the counter C4 is halved.
[0049]
As described above, encoding with a variable bit length is performed for each sample. The variable-length samples obtained by encoding are output as code data. Note that the variable length coding means 70 may have a function of separating the quantization noise component before performing the above processing. Specifically, a predetermined number of lower bits of each sample after the processing by the polarity processing means 60 are regarded as quantization noise components and separated. For example, when each sample is represented by 16 bits, in the present embodiment, the upper bits are separated into 12 bits and the lower bits are separated into 4 bits. This separation is basically performed to separate the thermal noise of a circuit used for digitizing an acoustic signal, such as an A / D converter. Therefore, lower bits considered to be thermal noise are separated. The degree to which the lower bits are separated depends on the characteristics of the sound source and the circuit used, but it is usually desirable to set the number of quantization bits to about 1/4. Therefore, in this case, 4 bits corresponding to 1/4 of 16 bits are separated as lower bits.
[0050]
Here, the state of data separation of upper bits and lower bits is schematically shown in FIG. In FIG. 7, H indicates upper bits or upper sample data, and L indicates lower bits or lower sample data. FIG. 7A shows sample data before separation. The variable length coding means 70 separates the sample data into upper sample data shown in FIG. 7B and lower sample data shown in FIG. Note that the sign bit included in the upper bits is directly included in the upper sample data and separated. As described above, when the quantization noise component is separated, the remaining upper bits are subjected to variable-length coding according to the flowchart shown in FIG. 6, and the lower bits are fixed as they are. Encoding is performed by length.
[0051]
The code data obtained as described above is stored as needed in a storage device such as a hard disk connected to a computer, and then stored in a format corresponding to a necessary storage medium.
[0052]
(Second embodiment)
Subsequently, an encoding device according to the second embodiment of the present invention will be described. FIG. 8 is a functional block diagram of an encoding device according to the second embodiment of the present invention. 8, components having the same functions as those in the configuration shown in FIG. 1 are denoted by the same reference numerals, and description thereof will be omitted. The difference from the first embodiment is that a signal flat part processing means 80 and a correlation frame detecting means 90 are added. In FIG. 2, the signal flat part processing means 80 has a function of detecting a flat part having a constant signal value for a sample sequence for each channel, and efficiently coding the same. After setting a predetermined section as a frame for each sample sequence, the correlation frame detection means 90 detects a correlation frame in which all the corresponding sample values are the same between frames, and detects the correlation frame backward (see FIG. It has a function of deleting the correlation frame located in the future. The apparatus shown in FIG. 8 is actually realized by a computer and a dedicated software program installed on the computer.
[0053]
Subsequently, the processing operation of the encoding device shown in FIG. 8 will be described. First, the mixed sound signal as described above is input from the time-series signal input means 10. Then, the inter-channel calculation means 50 performs a difference calculation process between the channels according to the procedure shown in the above figure. Subsequently, the sample sequence rearranging means 20 rearranges the sample sequence into the main sample sequence and the sub-sample sequence by the processing shown in FIG. After that, the lower fixed bit deleting means 30 deletes the lower bit of each sample.
[0054]
Next, the signal flat part processing means 80 performs processing of the signal flat part. The signal flat portion refers to a portion where the same signal level is continuous. In particular, it often appears in a silent part where the signal level is “0” and a saturated part where the absolute value of the signal level is the maximum. Silence occurs when the sound is actually silent or when the sound is not recorded very small, while saturation occurs during the signal recording and A / D conversion. Regardless of whether a silent portion, a saturated portion, or the other same signal level continues, the same signal level is continuously recorded for a predetermined time (a predetermined number of samples) in the signal flat portion. For this reason, this part is data that can be easily compressed. Specifically, the three values of the head time position of the signal flat portion, the number of samples following the same signal level, and the signal level (sample value) are recorded as signal flat portion data separately from the sample sequence of each channel. I do. The signal flat portion is deleted from the sample sequence of each channel. This is schematically shown in FIGS. 9A and 9B. FIG. 9A shows a sample sequence before the signal flat portion processing. In FIG. 9A, a shaded portion indicates a signal flat portion. By the processing of the signal flat part processing means 80, the signal flat part is deleted from the original sample sequence, as shown in FIG. 9B. However, in order to restore the original state at the time of decoding, the separated signal flat part is recorded as signal flat part data in a format as shown in FIG. 9C.
[0055]
As described above, the signal flat portion data is recorded for each signal flat portion with three attributes of the start time (sample number), the number of samples, and the sample value. Here, the head time is a time from the start position of the signal, and in the example of FIG. 9C, is recorded by the sample number from the head. If this sample number is divided by the sampling frequency, it will be converted to time. The sample number is information indicating how continuous the sample value continues. The end time of the signal flat portion may be recorded instead of the number of samples. The sample value indicates the digitized signal level. When represented by signed 16 bits, the maximum value is “32767” and the minimum value is “−32768”. That is, “0” indicates a silent part, and “32767” and “−32768” indicate a saturated part. However, the signal flat part processing means 80 does not unconditionally process the signal flat part. This is because the purpose of the present invention is to compress data, and it is meaningless if the signal flat portion data is larger than the reduction of the sample sequence. Therefore, only when the sample which becomes the signal flat portion continues for a predetermined number or more, the signal flat portion data is created and separated from the sample sequence of each channel.
[0056]
Subsequently, the correlation frame detection unit 90 sets a frame having a predetermined section length for the sample sequence of each channel, and performs comparison between the set frames. In the present embodiment, the frame length is fixed over the entire section from the start time to the end time of the sample sequence. Specifically, one frame has 512 samples. The correlation frame detection means 90 sets 512 samples from the head of the sample sequence of each channel as one frame, and obtains a correlation frame in which all samples match between frames. The specific procedure will be described with reference to the flowchart of FIG.
[0057]
First, the correlation frame detection unit 90 performs framing in units of a predetermined number of samples (step S31). In this embodiment, as described above, the frame length is set to 512 samples of the fixed length over the entire section from the start time to the end time of the sample sequence. As shown in FIG. 11A, the correlation frame detection means 90 sets 512 samples from the head of the sample sequence as one frame.
[0058]
Next, a search is made for a frame in which all sample values constituting each frame match. Specifically, as shown in FIG. 11B, first, a temporally last frame among the set frames is set as a target frame for searching for a correlation frame. Next, within the predetermined search range, a sample having the same value as the value of the first sample of the target frame is searched while going back in time (step S32). For example, as shown in FIG. 12A, it is assumed that the target frame is composed of 512 samples of mT to mT + 511. In this case, first, a sample that is the same as the sample value e (mT) of the first sample mT of the target frame is searched for. The sample mT-1 and the sample mT-2 are sequentially searched. In FIG. 12, m indicates the m-th frame from the beginning, and T indicates the frame length (512 samples in this embodiment).
[0059]
When a matching sample t is found (step S33), it is next compared whether the next sample t + 1 of the sample t matches the second sample mT + 1 of the target frame. In this way, comparison between subsequent samples is performed as long as the values of the samples match (step S34). In step S34, the process is repeated as long as the values of e (t + p) and e (mT + p) match. For example, in the example shown in FIG. 12B, since e (t) to e (t + 8) match with e (mT) to e (mT + 8), the process of step S34 is continued with p = 9. Will be done. If all of e (t + p) and e (mT + p) from p = 0 to p = 511 match (step S35), the sample sequence is set as a correlation frame for the target frame, and the first sample number of the correlation frame and the target frame Is recorded as frame correlation data in association with the first sample number, and the target frame is deleted from the original sample sequence (step S36). If the sample does not match all the samples of the target frame, a search is further performed in time to determine whether there is a sample whose value matches the first sample of the target frame. If there is no matching correlation frame even after going back by the predetermined number of samples, the search for the correlation frame related to the target frame is stopped, and the search for the correlation frame is performed using the frame immediately before the target frame as a new target frame. When the processing for one target frame is completed, the process returns to step S32, and the processing is continued with the immediately preceding frame as a new target frame (step S37). In this way, the detection processing of the correlation frame is performed with all the frames except the frame located near the head time of the time series signal as the target frame.
[0060]
Looking at the entire sample sequence, assuming that a correlation frame corresponding to the target frame is detected as shown in FIG. 11C, the target frame is deleted as shown in FIG. 11D. At this time, frame correlation data as shown in FIG. 11 (e) is recorded so that it can be completely restored at the time of decoding. As shown in FIG. 11E, the head sample number of the target frame and the head sample number of the correlation frame are recorded in the frame correlation data in association with each other.
[0061]
The preferred embodiment of the present invention has been described above, but the present invention is not limited to the above embodiment, and various modifications are possible. For example, in the above-described embodiment, the case where the appearance ratio of the main sample and the sub-sample is 1 to 1 has been described, but another ratio may be used. In the above embodiment, the sound signal created by first interpolating the sampled signal at 48 kHz to 96 kHz was used, so that the appearance ratio of the main sample and the sub-sample was 1: 1. When the sampled sample is interpolated to 96 kHz, the appearance ratio of the main sample and the sub-sample becomes 1: 3. When dealing with such an acoustic signal, the sample sequence rearranging means 20 may divide each sample into one main sample sequence and three sub-sample sequences in order.
[0062]
【The invention's effect】
As described above, according to the present invention, for a sample sequence constituting a time-series signal, a main sample sequence created by recording and a sub-sample sequence obtained by interpolating the main sample sequence are described. Separating and performing a sample rearrangement to convert the value of each sub-sample in the sub-sample sequence into a difference between the average value of the nearby main sample and the value of the sub-sample, the main sample sequence is converted. A linear prediction error is calculated for each sub-sample sequence, the values of the main sample sequence and the sub-sample sequence are converted into prediction error values, and the main sample sequence and the sub-sample sequence converted into the prediction error values are variable-length. Since the encoding is performed by using, the digitally edited high-definition audio work data can be efficiently reversibly compressed.
[Brief description of the drawings]
FIG. 1 is a functional block diagram showing an apparatus for encoding a time-series signal according to a first embodiment of the present invention.
FIG. 2 is a diagram showing a state of sample rearrangement by a sample sequence rearrangement unit 20;
FIG. 3 is a flowchart showing processing by a prediction error conversion unit 40;
FIG. 4 is a flowchart showing processing by an inter-channel calculation means 50;
FIG. 5 is a diagram showing a state of conversion of a bit configuration by a polarity processing unit 60;
FIG. 6 is a flowchart showing a process performed by a variable-length encoding unit 70;
FIG. 7 is a diagram showing a state of data separation of upper bits and lower bits.
FIG. 8 is a functional block diagram illustrating a time-series signal encoding device according to a second embodiment of the present invention.
FIG. 9 is a diagram showing a state of processing by a signal flat part processing means 80;
FIG. 10 is a flowchart illustrating a process performed by a correlation frame detection unit 90;
FIG. 11 is a diagram showing a state of an entire time-series signal by processing of a correlation frame detecting means 90.
FIG. 12 is a diagram showing a state of samples compared by the processing of the correlation frame detection means 90.
[Explanation of symbols]
10 time-series signal input means
20 ... sample row rearrangement means
30... Lower fixed bit deleting means
40 ... Prediction error conversion means
50: Channel calculation means
60 ... polarity processing means
70 ... variable length coding means
80 ... Signal flat part processing means
90 ... correlation frame detecting means

Claims (11)

A coding apparatus that compresses the amount of information so that all the sample sequences can be reproduced with respect to a time-series signal including a time-series sample sequence,
For the sample sequence, a main sample sequence obtained by sampling a recording signal and a sub-sample sequence obtained by interpolating the main sample sequence are separated, and each sub-sample sequence in the sub-sample sequence is separated. Sample sequence rearrangement means for converting the value of the sample into a difference between the average value of the nearby main sample and the value of the subsample;
Prediction error conversion means for calculating a linear prediction error for each of the main sample sequence and the converted sub-sample sequence, and converting the values of the main sample sequence and the sub-sample sequence into prediction error values,
The main sample sequence converted to the prediction error value, a variable length encoding means for encoding the sub-sample sequence with a variable length,
An encoding device for a time-series signal, comprising: In claim 1,
The sample sequence rearrangement means, when a value of an even-numbered sample sequence of the sample sequence is close to the average value of the preceding and following odd-numbered sample sequence, the odd-numbered sample is a main sample sequence, an even-numbered sample sequence. To separate the sample located at the
When the value of an odd-numbered sample in the sample sequence is close to the average value of the preceding and following even-numbered sample sequences, the even-numbered sample is a main sample sequence, and the odd-numbered sample is a sub-sample. An encoding device for a time-series signal, which is separated into columns. In claim 1,
A low-order fixed bit deletion unit that deletes a predetermined number of lower-order bit components from the main sample sequence and the sub-sample sequence processed by the sub-sample sequence conversion unit; Wherein the prediction error conversion means performs processing on the sample sequence of (1). In claim 1,
When the sample sequence is composed of a plurality of channels having a plurality of values at the same time, the difference between the samples at the same time between the channels is calculated, and the sample sequence of any channel is calculated as the calculated difference value. A time-series signal encoding apparatus, further comprising an inter-channel calculation means for updating the time-series signal. In claim 1,
In the sample sequence, a sample in which the value of the sample is continuously the same is extracted and deleted from the sample sequence, and the start time position of the deleted sample, the number of samples, and the sample value are calculated. An encoding apparatus for a time-series signal, comprising signal flat part processing means for encoding two values at a stage preceding the prediction error conversion means. In claim 1,
If a frame having the same content as a frame composed of a predetermined number of sample sequences exists in the past in the sample sequence, the frame located in the future is deleted, and the start time position of both frames, An apparatus for encoding a time-series signal, comprising: a correlation frame detecting means for encoding the number of samples forming a frame, at a stage preceding the prediction error converting means. In claim 1,
The prediction error conversion means calculates a plurality of prediction error value candidates for the main sample sequence and the sub-sample sequence from a temporally past sample sequence based on a plurality of prediction calculation formulas. A time-series signal encoding apparatus for selecting a prediction error value to be encoded from the data. In claim 7,
An apparatus for encoding a time-series signal, wherein the linear coefficients of the plurality of prediction formulas are updated every predetermined number of samples. In claim 1,
Polarity processing means for shifting the whole of the absolute value of the sample value converted to the prediction error by one bit higher and inserting a sign into the least significant one bit is provided before the variable length coding means. A time-series signal encoding apparatus characterized by the above-mentioned. In claim 1,
The variable-length encoding means encodes lower-order bit components as they are among bit components of each sample converted to the prediction error value, and changes bit components for the remaining upper-order bit components. A time-series signal encoding apparatus for performing encoding by using a time-series signal. A recording medium in which, for a given time-series signal, code data output by the time-series signal encoding device according to any one of claims 1 to 10.

Images