JPH11261422A - Method and device for bit length extension - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a bit length extension method that can generate lower bit data suitable to any signal patterns and does not need to store the huge amount of lower bit data in accordance with each signal pattern in performing bit length extension of digital data. SOLUTION: A pattern of input data is detected and the number (b) of samples P to Q where the same value succeeds is investigated. When the (b) is above a specified number, the first sample P is converted into a value obtained by adding -0.5 (LSB) to the data value. The last sample Q is converted into a value obtained by adding a value most proximate to +0.5 (LSB) but not more than +0.5 (LSB) to the data value.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] 1. Field of the Invention [0002] The present invention relates to a bit length extension method and a bit length extension device for extending the bit length of digital data, especially digital audio data.

[0002]

2. Description of the Related Art In digital audio in which an analog signal is converted into a digital signal and recorded, the current bit length is 16 to 16 bits.
A high-precision A / D converter of about 20 bits is used. However, when these technologies were introduced, A / D converters having a bit length of about 13 to 14 bits were used, and music sources recorded using these devices were recorded on media such as CDs and released. In this case, the upper 13 bits of 16 bits are used, and the lower 3 bits are zero data. For example, the data structure of 16 bits is employed, and the data is substantially 13 bit data. It was meaningful.

[0003] High-precision D / A converters having a bit length of about 20 bits have been widely used. To make effective use of this, 16-bit data recorded on a CD
There is a need to extend the bit length to about 20 bits in the D player.

In view of these circumstances, Japanese Patent Application Laid-Open No. 9-980
No. 92, etc., an LPF (Low P
There is a technique of removing unnecessary harmonic (high-frequency) distortion using an assembler filter and extending the bit length of N-bit digital audio data to M-bit (M> N) digital audio data.

However, the conventional bit length extension technology needs to be mounted on hardware to perform real-time processing. Therefore, in order to generate lower bits, several types of moving average filter outputs having different amplitude-frequency characteristics are used as signal patterns. Or a method in which several kinds of lower-order bit patterns are stored in a ROM (Read Only Memory) or the like and switched according to a signal pattern. Although these methods have the advantage that the hardware scale can be reduced and the processing speed is high (real-time processing is possible), since only LPFs and low-order bit patterns that match several types of signal patterns are provided, all the signal patterns are used. It was not possible to generate optimal lower-order bit data.

[0006] In the method of storing lower bit data corresponding to each signal pattern in a ROM or the like and sequentially switching the lower bit data, lower bit data corresponding to all signal patterns are stored. Lower bit data optimal for a signal pattern can be generated (added). However, in this method, in order to generate optimal lower bit data for all signal patterns, it is necessary to store an enormous amount of lower bit data in a ROM or the like.

[0007]

SUMMARY OF THE INVENTION The present invention has been made in view of the above circumstances, and when extending the bit length of digital data, suitable low-order bit data is applied to any signal pattern. Can be generated,
Moreover, it is an object of the present invention to provide a bit length extension method and a bit length extension device which do not need to store an enormous amount of lower bit data according to each signal pattern.

[0008]

According to the first aspect of the present invention, a first digital data having an effective bit number N is extended to a second digital data having an effective bit number M (M> N). In the extension method, a signal pattern of the first digital data is detected, and when there are samples in which data of the same value continues for a predetermined number or more, the data of the first sample and the last sample of the continuous samples are added to the data. On the other hand, data correction for increasing or decreasing the value of 1/2 ^MN of the value of the least significant bit in the number N of effective bits by an integral multiple of 2 ^MN-1 or less is performed, and the validity of the first digital data is corrected. It is characterized by extending the number of bits.

According to a second aspect of the present invention, in the bit length extending method according to the first aspect, when the data of the continuous samples is a local maximum value or a local minimum value, a central portion of the continuous samples is provided. Is 1/2 of the value of the least significant bit in the number N of effective bits.
^The present invention is characterized in that data correction for increasing or decreasing the value of ^{MN by} an integral multiple of 2 ^MN-1 or less is performed to extend the number of effective bits of the first digital data.

According to a third aspect of the present invention, in the bit length extension method according to the first or second aspect, the data of the continuous samples is subjected to a spline interpolation method based on the data-corrected samples. With respect to the data of the middle sample of the data-corrected samples in the continuous samples, the value of 1/2 ^MN of the value of the least significant bit in the effective bit number N is 2
^The present invention is characterized in that data correction for increasing or decreasing by an integral multiple of ^MN-1 or less is performed to extend the number of effective bits of the first digital data.

According to a fourth aspect of the present invention, in the bit length extension method according to any one of the first to third aspects,
When the samples for which the data correction is performed are continuous, a value to be increased or decreased so that the corrected data does not have the same value is determined.

According to a fifth aspect of the present invention, in the bit length extension method according to any one of the first to third aspects,
The value for increasing (or decreasing) the data of the sample to be subjected to the data correction is a half of the value of the least significant bit in the number N of effective bits, and the value for decreasing (or increasing) the value of the validity is N. The value is (2 ^MN−1 −1) times the value of 倍^MN of the value of the least significant bit in the number of bits N.

According to a sixth aspect of the present invention, the number of effective bits N
Pattern detecting means for detecting the signal pattern of the first digital data, and data correcting means for performing data correction on the sample data in accordance with the output of the pattern detecting means, wherein the number of effective bits M (M> N) A bit length extension device for extending to a second digital data, wherein the data correction means is configured to add 1 to the value of the least significant bit in the number N of effective bits with respect to data of a sample in which data of the same value continues. / 2 ^MN to perform data correction to increase or decrease by an integral multiple of 2 ^{MN -1} or less, wherein the pattern detection means sets the data of the sample of interest to a value different from the data of the immediately preceding sample. In the case where there is a change point where data of the same value continues after that, the count value b obtained by counting the number of samples in which the data of the same value continues is counted. It is characterized in that force.

According to a seventh aspect of the present invention, in the bit length extending apparatus according to the sixth aspect, the data correction means comprises:
When the count value b is 3 or more, the data of the sample at the transition point is increased by an integral multiple of 2 ^MN-1 or less of 1/2 ^MN of the value of the least significant bit in the number of effective bits N or Data reduction is performed, and the data of the sample after (b-1) samples from the sample at the transition point is ^{added to} the value of 2 ^MN− the value of 1/2 ^MN of the value of the least significant bit in the number N of effective bits. Data correction for increasing or decreasing by an integral multiple of ¹ or less is performed.

According to an eighth aspect of the present invention, in the bit length extension device according to the sixth or seventh aspect, the data correction means is configured to determine that the count value b of consecutive samples is 5 or more, and If the data of the sample to be performed is a local maximum value or a local minimum value, the data of one or two samples at the center is ^replaced by 2 ^MN−1 of the value of 1/2 ^MN of the value of the least significant bit in the effective bit number N. It is characterized by increasing or decreasing by the following integral multiple.

According to a ninth aspect of the present invention, there is provided the bit length extending apparatus according to any one of the sixth to eighth aspects,
The data correction means, by a spline interpolation method based on the sample data corrected for the data of the continuous sample, for the data of the intermediate sample of the data corrected sample in the continuous sample, Data correction is performed to increase or decrease by an integral multiple of 2 ^MN-1 or less of 1/2 ^MN of the value of the least significant bit in the effective bit number N.

According to a tenth aspect of the present invention, in the bit length extension method according to any one of the sixth to ninth aspects, the data correction means is configured to perform the data correction when the sample to be subjected to the data correction is continuous. It is characterized in that a value to be increased or decreased is determined so that the obtained data does not have the same value.

According to an eleventh aspect of the present invention, in the bit length extension method according to any one of the sixth to ninth aspects, the data correction means increases (or decreases) the data of the sample to be corrected. The value to be decreased is a half of the value of the least significant bit in the number N of effective bits, and the value to be decreased (or increased) is one half of the value of the least significant bit in the number N of effective bits. it is characterized in that it is (2 ^MN-1 -1) times the value of the value of the ^MN.

[0019]

FIG. 1 is a block diagram showing an example of an embodiment of a bit length extending apparatus according to the present invention. In the figure, reference numeral 1 denotes a pattern detection unit, and 2 denotes a data correction unit. The pattern detecting means 1 detects a sequential pattern of a sample of input data. Here, the input data has a bit length of N. In this embodiment, the pattern detecting means 1
Detects a change point by sequentially looking at a change state of data of time-series samples of an input signal. If the absolute value of the difference between the data of the sample of interest and the data of the sample one sample before is 1 SLB or more, the sample of interest becomes the sample of the change point. When the change point is detected, the number of samples having the same value is continuously counted later, and a count value b, which is the number of samples having the same value, is output. Therefore, the count value b can be said to be the number of samples in which the same value continues including the change point.

FIG. 2 is an explanatory diagram of a pattern detecting method in the pattern detecting means. The black circles indicate the position in the time series of each sample and the level which is the size of the data.
The sample P at the time point n differs from the sample O at the time point (n-1) by 1 SLB or more of the N-bit data (in this specific example, the change is 1 SLB). Therefore, the sample P at the time point n is a sample at the change point.
The samples at each time point of (n + 1), (n + 2), (n + 3) and (n + 4) have the same value as the sample at the time point n.
Therefore, for the sample P at the change point, the pattern detection means outputs b = 5 as the count value b from the sample P as a starting point to the last sample Q in which the same value continues. The sample R at the (n + 5) time point is a change point because the data of the sample R is different from the value of the sample Q at the (n + 4) time point. However, the sample S at the time (n + 6) after the sample R at the time (n + 5) is also a change point, and (n + 5)
5) Since there is no sample having the same value following the sample R at the time point, the pattern detecting means
= 1 is output. At other points than the change point, the pattern detection means outputs b = 1. At points other than the transition point, b = 0 may be output, or the count value may not be output.

Returning to FIG. Data correction means 2
Is obtained by performing data correction described below on samples having the same value in succession to obtain a bit length M (M>
N) is output.

An example of an embodiment of a data correction method in the data correction means will be described. In the data correction method according to this embodiment, data correction is performed when the output b of the pattern detection means meets the following condition.

That is, when the output b of the pattern detecting means is 3 or more with respect to the digital data having the number of effective bits N, successive samples of data having the same value are subjected to level conversion according to the following conditions. Data correction,
The data is corrected based on the value calculated by the spline interpolation method, and digital data having an effective bit number M (M> N) is output. If it is a change point and the output b of the pattern detecting means is 2 or less, the data correcting means outputs data of the effective bit number M without changing the data value of the input signal. That is, (M−N) 0s are added to the lower part of the input data and output.

FIGS. 3 to 9 are explanatory diagrams of an embodiment of level conversion. FIG. 3 is a diagram similar to FIG. 2, wherein a black circle indicates a time-series position of each sample and a level which is a data size. Open circles indicate data of the sample subjected to the data correction.

FIGS. 3 and 4 show a case where data correction is performed for the first sample and the last sample of a sample in which the same value continues for a predetermined number of times.

FIG. 3 shows a case where data tends to increase on the time axis. The sample P, which is a change point, is converted to a value obtained by adding -0.5 (LSB) to the value of the data. Also, counting from the change point, excluding the change point, (b−
1) Regarding the data of the sample Q after the sample, the value is the most +0.5 (LSB) less than +0.5 (LSB).
Is converted to a value obtained by adding a value close to.

FIG. 4 shows a case where the data is decreasing on the time axis. For sample P, which is a change point, the value of the data is the most +0.5 (LSB) less than +0.5 (LSB).
The value is converted to a value obtained by adding a value close to (LSB). Also, counting the data excluding the change point, the data of the sample Q after (b-1) samples from the change point has a value of -0.5 (L
(SB).

As described above, when the value of the first sample of the consecutive samples is larger than the value of the immediately preceding sample, the value of the first sample is reduced by 0.5 (LSB). If the value of the sample is smaller than the value of the immediately preceding sample, 0.5 (LS
B) is converted to a value increased by a value closest to 0.5 (LSB) less than B). When the value of the sample after the last sample of the last sample of the continuous sample is reduced, the value is converted to a value reduced by 0.5 (LSB), If the value of the next sample after the value of is increased, 0.5 (LS
B) is converted to a value increased by a value closest to 0.5 (LSB) less than B).

The data correction method described with reference to FIGS.
In a bit length extension method for extending first digital data having an effective bit number N to second digital data having an effective bit number M (M> N), a signal pattern of the first digital data is detected, and When there is a sample in which the data of the value is continuous for a predetermined number or more, the data of the first sample of the continuous samples increases from the data of the previous sample by increasing the data of the first sample. If you are, the effective in number of bits N by 2 ^MN-1 an integer multiple of the 1/2 ^MN least significant bits of the value decreasing, the first sample data before one sample When the number is smaller than the data, the value of the least significant bit in the effective bit number N is 1/2 ^MN.
Is corrected by an integer multiple of 2 ^MN-1 or less, and the data of the last sample in the continuous sample is replaced by the data of the sample one sample after the last sample. If the number of data is larger than that of the last data, it is increased by an integer multiple of 2 ^MN-1 or less of a value of 1/2 ^MN of the value of the least significant bit in the number N of effective bits, and When the data of the sample one sample after the sample is smaller than the data of the last data, 2 ^MN-1 or less of 1/2 ^MN of the value of the least significant bit in the effective bit number N It can be said that this is a method of extending the effective number of bits of the first digital data by performing data correction to reduce the number by an integral multiple of.

FIGS. 5 to 8 show the case where successive samples having the same value show the maximum value or the minimum value.

FIG. 5 shows a case where successive samples having the same value indicate the maximum value and b is an even number. In the example of FIG. 5, the number b of the samples having the same value successively is b = 6, and for the first and last samples, FIGS.
The data conversion described above is performed. That is, the data of the sample P at the transition point is converted into a value obtained by adding -0.5 (LSB) to the value, and the data is counted excluding the transition point. , The value is -0.5 (L
(SB). The data of the samples T and U after ((b / 2) -1) and (b / 2) samples from the change point are counted by excluding the change point, and the value thereof is added by +0.5.
The value is converted to a value obtained by adding a value closest to +0.5 (LSB) less than (LSB). That is, the conversion for increasing the data value is performed on the two samples at the center of the samples in which the same value is an even number and continues.

FIG. 6 shows a case where successive samples having the same value show the maximum value and b is an odd number. In the example of FIG. 6, the number b of samples having the same value continues is b = 5. For the first and last samples, FIGS.
The data conversion described above is performed. That is, the data of the sample P at the transition point is converted into a value obtained by adding -0.5 (LSB) to the value, and the data is counted excluding the transition point. , The value is -0.5 (L
(SB). Also, counting the data excluding the change point, the data of the sample T after (b-1) / 2 samples from the change point is represented by the value closest to +0.5 (LSB), which is less than +0.5 (LSB). Is converted to the value obtained by adding. That is, a conversion for increasing the data value is performed on one sample at the center of a sample in which the same value continues in an odd number.

FIG. 7 shows a case where successive samples having the same value indicate the minimum value and b is an even number. In the example of FIG. 7, the number b of samples having the same value successively is b = 6, and for the first and last samples, FIGS.
The data conversion described above is performed. That is, the data of the sample P at the change point is converted into a value obtained by adding a value less than +0.5 (LSB) and closest to +0.5 (LSB) to the value,
Counting excluding the transition point, the data of the sample Q after (b-1) samples is converted into a value obtained by adding a value less than +0.5 (LSB) and closest to +0.5 (LSB) to that value. .
Also, counting from the changing point, counting from the changing point, ((b / 2)
The data of the samples T and U after the -1) and (b / 2) samples are converted into a value obtained by adding -0.5 (LSB) to the value. That is, the conversion for reducing the data value is performed on the two samples at the center of the samples in which the same value is an even number and continues.

FIG. 8 shows a case where successive samples having the same value show the minimum value and b is an odd number. In the example of FIG. 8, the number b of samples having the same value is b = 5. For the first and last samples, b = 5.
The data conversion described above is performed. That is, the data of the sample P at the change point is converted into a value obtained by adding a value less than +0.5 (LSB) and closest to +0.5 (LSB) to the value,
Counting excluding the transition point, the data of the sample Q after (b-1) samples is converted into a value obtained by adding a value less than +0.5 (LSB) and closest to +0.5 (LSB) to that value. .
Also, counting from the changing point, counting from the changing point, (b-1) /
The data of sample T after two samples is added to the value of −0.0.
Converted to a value obtained by adding 5 (LSB). That is, the conversion for reducing the data value is performed on one sample at the center of the sample in which the same value is an odd number consecutive sample.

In the embodiment described with reference to FIGS. 3 to 8, as described above, the value to be reduced is 0.5 (LS
B), and the value to be increased was a value less than 0.5 (LSB) and closest to 0.5 (LSB). The value to be decreased and the value to be increased may both be 0.5 (LSB).
However, both the decreasing value and the increasing value are 0.5
In the case where (LSB) is set, when samples to be corrected continuously exist, for example, when the same sample continues after the sample R in FIG.
Is converted to a value reduced by 0.5 (LSB).
Is corrected to increase by 0.5 (LSB). Then, the samples Q and R subjected to the data correction have the same value, and the data correction results in successive samples of the same value, which is not preferable from the viewpoint of reducing the quantization noise. Therefore, when samples for which data correction is performed are continuous, it is preferable to determine a value that increases or decreases so that the corrected data does not have the same value. In this embodiment, for this reason, the value to be reduced is set to 0.5
(LSB), and the value to be increased is set to a value closest to 0.5 (LSB), which is less than 0.5 (LSB). Conversely, the value to be decreased is set to a value closest to 0.5 (LSB) less than 0.5 (LSB), and the value to be increased is set to:
It may be 0.5 (LSB). Further, both the value to be decreased and the value to be increased may be a value less than 0.5 (LSB) and closest to 0.5 (LSB). Note that 0.5
An explanation will be added for a value less than (LSB) and closest to 0.5 (LSB). For example, when the data length of the input data is extended by 3 bits, 1 (LSB) of the input data is used.
Is divided into ８. In this case, 0.5
The value closest to 0.5 (LSB) less than (LSB) is 3
/ 8 (LSB). Note that 1/8 (LSB) is 1 (LSB) of the output data whose bit length has been extended.

Regarding the value to be decreased and the value to be increased, assuming that the input data is obtained by rounding the analog signal and quantizing it, the magnitude V of the analog signal quantized to a certain level is: For that level L,
Since it can be said that L−0.5 (LSB) ≦ L <L + 0.5 (LSB), the maximum value of the data correction is 0.5
(LSB) is reasonable. Further, as described above, when considering the quantization by rounding, the value to be decreased is set to 0.5 (LSB), and the value to be increased is smaller than 0.5 (LSB) and is closest to 0.5 (LSB). It can be said that the value is more rational in view of the range of the analog data.

Further, a value for data conversion of the sample to be corrected may be determined from a spline curve passing through several points before and after the sample to be corrected by using a spline interpolation method described later. Therefore, the value to be decreased and the value to be increased for the value of the data to be corrected are 0.5 (LS
B) or a value less than 0.5 (LSB) and closest to 0.5 (LSB).

When the number b of consecutive samples having the same value is small, a considerable effect can be expected by the data correction described above. But to be more effective,
If the number b of consecutive samples of the same value is large, a spline interpolation method can be employed. As an example, for the data before and after the sample P at the change point shown in FIG.
FIG. 9 shows a state in which the data correction described in FIGS. 3 to 8 has been performed. This data correction is performed by converting the sample P in FIG. 2 into a value obtained by adding -0.5 (LSB) to the value, counting the sample P excluding the sample P, and counting the sample Q after (b-11) samples from the sample P. Is added to that value by +0.5 (LS
This is a state in which the value is converted to a value obtained by adding a value closest to +0.5 (LSB) less than B). Regarding the samples T, U, and V between the samples P and Q, the data correction method described with reference to FIGS. 3 to 8 does not perform data correction, and the same value remains continuous.

FIG. 10 shows a case where the data-corrected sample P is used as a starting point and the data-corrected sample Q is used as an ending point.
Samples T, U, and V in the meantime are calculated as data of the effective bit number M by the spline interpolation method, and the input data of the effective bit number N is replaced with the data. By this replacement, the data of the samples T, U, and V in the middle of the data-corrected samples are increased or decreased from the original data, and the data is corrected.

Although the order of the spline interpolation is not limited,
If the degree is increased, the calculation becomes complicated and it takes time. A case where cubic spline interpolation is applied will be described. Cubic spline interpolation (function)
, It is necessary to give the slope (boundary condition) between the start point and the end point as an initial value. Thus, in the present bit length extension device, when the starting point is f (x) and the ending point is f (x + a), the boundary condition f ′ (x) = f (x) −f (x−1) of the starting point, Boundary condition f '(x + a) = f (x + a + 1) -f
(X + a).

In using the spline interpolation method, data obtained by correcting the data of the first and last samples of consecutive samples having the same value is used as a start point and an end point, and a sample at the center of the continuous sample, for example, When the number of samples b is odd, one sample at the center may be used as the point used for the spline interpolation method, and when the number of samples b is even, two samples at the center may be used. Of course, as described with reference to FIGS. 5 to 8, when data correction is performed on the sample in the center, the sample on which the data correction has been performed may be used in the spline interpolation method.

A flowchart of the processing in the above-described embodiment is shown in FIG. 11 and will be described. First, in S1, when input data having a bit length N is given, samples are sequentially taken in. In S2, it is determined whether or not the fetched data is EOF (data notifying the end of the data). If the input data is EOF, end processing is performed. If the input data is not EOF, the process proceeds to the next processing.
In S3, the pattern of the input data is detected, and it is determined whether or not the sample to be processed is a change point.
The sample to be processed is a sample that is a predetermined number of samples, for example, 50 samples before the currently acquired sample. If the sample to be processed is a change point, in step S4, if b = 1, no data correction is performed, and in step S7, (M−
N) 0s are added and the data of bit length M
Is output. If the input data has other conditions, predetermined level conversion and cubic spline interpolation processing are performed in S5, converted to data of bit length M by data correction, and output in S6. In this way,
Suitable correction can be performed, and it is not necessary to store lower-order bit data in a ROM or the like, and the bit length can be extended. In S2, when the sample acquired is EOF, the end processing is performed at S3 and S5 when the sample corresponding to EOF is acquired. When the processing of the unprocessed sample is completed, the flow ends. Unprocessed data may be discarded depending on the content of the data.

A specific example will be described. FIG. 12, FIG.
Shows an example of the time course of an analog signal. When this is converted into a digital signal using 13-bit bit length data, it is quantized as shown in FIGS. 13 and 17, and the level value of the infinite number of digits represented by the analog signal and the finite Level value (13-bit 1L
An error occurring between the signal and the corresponding SB unit becomes quantization noise (quantization noise). When the number of bits of the digital signal is increased, the quantization noise is reduced, but it can be said that the quantization noise is generated even when the number of bits is high.

In the present invention, the bit length is extended. In this example, a digital signal having 13 effective bits is converted into a digital signal having 16 effective bits.

The conventional bit length extension device has one LSB (1
A portion that has become a rectangular wave due to a level change in units (in 3-bit data) is detected, and low-pass filter (LPF) processing is performed, and the level is reduced to 1/8 of 1 LSB (in 13-bit data). Is converted into a digital signal in units of 1 LSB (in 16-bit data), and as shown in FIGS. 14 and 18, the level change becomes smooth and the distortion component is reduced. However, the conventional bit length expansion device detects a predetermined pattern (a portion where the same level is continuous for a plurality of samples), has a filter corresponding to a predetermined number of patterns, and uses the number of filters to set the bit length. It is a method of expanding. For example, when the same level has two types of LPFs corresponding to a pattern in which three samples and ten samples are continuous, even when the same level is continuous for five samples, the bit length is extended using the LPF for three samples. I have. Thereby, as shown in FIG. 14 and FIG.
A portion where the same level is continuous for three samples is generated, and a portion where the same level is continuous for five samples is a smooth signal near the change point, but is not processed near the center and is not a smooth signal.

In the case of the bit length extension according to the present invention, since the same level is subjected to the bit length extension process in conformity with the number of consecutive samples, as shown in FIGS. In addition, the bit length extension processing according to each pattern can be performed on the portion where the same level continues for five samples, and a digital signal that is closer to the original analog signal can be obtained.

[0047]

As is apparent from the above description, according to the present invention, it is possible to generate low-order bit data suitable for any signal pattern, and furthermore, it is possible to generate low-order bit data suitable for each signal pattern. RO with huge amount of lower bit data
The bit length can be extended without having to store it in M or the like.

[Brief description of the drawings]

FIG. 1 is a block diagram showing an example of an embodiment of a bit length extension device according to the present invention.

FIG. 2 is an explanatory diagram of a pattern detection method in a pattern detection unit.

FIG. 3 is an explanatory diagram of an embodiment of level conversion.

FIG. 4 is an explanatory diagram (continued) of an embodiment of level conversion;

FIG. 5 is an explanatory diagram (continued) of an embodiment of level conversion.

FIG. 6 is an explanatory diagram (continued) of the embodiment of the level conversion.

FIG. 7 is an explanatory diagram (continued) of the embodiment of the level conversion.

FIG. 8 is an explanatory diagram (continued) of the embodiment of the level conversion.

FIG. 9 is an explanatory diagram (continued) of the embodiment of the level conversion.

FIG. 10 is an explanatory diagram (continued) of the embodiment of the level conversion.

FIG. 11 is a flowchart of an example of a process according to the embodiment of the data correction of the present invention.

FIG. 12 is a diagram illustrating an example of a time course of an analog signal.

13 is an explanatory diagram in which the analog signal of FIG. 12 is quantized.

FIG. 14 is an explanatory diagram in which the digital signal of FIG. 13 is extended in bit length by a conventional method.

FIG. 15 is an explanatory diagram in which the digital signal of FIG. 13 is extended in bit length by the method of the present invention.

FIG. 16 is a diagram showing another example of a time course of an analog signal.

17 is an explanatory diagram in which the analog signal of FIG. 16 is quantized.

18 is an explanatory diagram in which the digital signal of FIG. 17 is extended in bit length by a conventional method.

19 is an explanatory diagram in which the digital signal of FIG. 17 is extended in bit length by the method of the present invention.

[Explanation of symbols]

 1 ... pattern detection means, 2 ... data correction means.

Claims (11)

[Claims] 1. A bit length extension method for extending first digital data having an effective bit number N to second digital data having an effective bit number M (M> N), wherein a signal pattern of the first digital data is provided. Is detected, and if there is a sample in which data of the same value continues for a predetermined number or more,
With respect to the data of the first sample and the last sample of the consecutive samples, the value is increased by an integer multiple of 2 ^MN−1 or less of the value of 1/2 ^MN of the value of the least significant bit in the number of effective bits N or A bit length extension method, wherein the number of effective bits of the first digital data is extended by performing data correction to reduce the number of effective bits. 2. When the data of the continuous samples is a local maximum value or a local minimum value, the data of the central sample of the continuous samples is replaced with the effective bit number N.
2 values of 1/2 ^MN least significant bits of the values in
^2. The bit length extension method according to claim 1, wherein data correction is performed to increase or decrease by an integral multiple of ^MN-1 or less to extend the number of effective bits of the first digital data. 3. A method according to claim 1, wherein the data of the continuous sample is interpolated by a spline interpolation method based on the data-corrected sample. , By performing data correction to increase or decrease by an integral multiple of 2 ^MN-1 or less of 1/2 ^MN of the value of the least significant bit in the effective bit number N,
3. The method according to claim 1, wherein the number of effective bits of the digital data is extended. 4. The method according to claim 1, wherein when the samples for which the data correction is performed are continuous, a value to be increased or decreased so that the corrected data does not have the same value is determined. The described bit length extension method. 5. A value for increasing (or decreasing) the data of the sample to be subjected to the data correction is a value which is １ ／ of the value of the least significant bit in the effective bit number N, and is decreased (or increased). The value is a value of (2 ^{MN -1} ) of 1/2 ^MN of the value of the least significant bit in the effective bit number N.
4. The method according to claim 1, wherein the value is a multiple of -1) times.
The bit length extension method according to any one of the above. 6. A data processing apparatus comprising: pattern detection means for detecting a signal pattern of first digital data having an effective bit number N; and data correction means for performing data correction on sample data in accordance with an output of the pattern detection means. In the bit length extension device for extending to the second digital data having the number of bits M (M> N), the data correcting means may determine the maximum number of the effective bits N with respect to the data of the sample in which data of the same value continues. 1/2 of the value of the lower bit
^Performing data correction to increase or decrease by an integral multiple of 2 ^MN-1 or less of the value of ^MN , wherein the pattern detecting means is such that the data of the sample of interest is different from the data of the immediately preceding sample; A bit length expansion device that outputs a count value b obtained by counting the number of samples in which the same value data continues when the same value data continues at a transition point thereafter. 7. The data correction means, when the count value b is 3 or more, includes:
1 / (1) of the value of the least significant bit in the number of effective bits N
By an integral multiple of 2 ^MN-1 below the value of 2 ^MN performs increase or decrease the data correction, and, from the sample point of change in (b-1) data of the sample after sample, most of the effective number of bits N 1/2 ^MN of lower bit value
7. The bit length extension device according to claim 6, wherein data correction is performed to increase or decrease by an integer multiple of 2 ^MN-1 or less of the value of. 8. The data correction means, when the count value b of a continuous sample is 5 or more and the data of the continuous sample is a local maximum value or a local minimum value, the 1 or 2 of the central portion The sample data is set to M ^MN of the value of the least significant bit in the number N of effective bits.
The bit length extending apparatus according to claim 6 or 7, wherein the value is increased or decreased by an integral multiple of 2 ^MN-1 or less of the value of? 9. The data correction means according to claim 1, wherein the data correction means comprises a spline interpolation method based on the data-corrected samples of the continuous samples, and the spline interpolation method of the intermediate samples of the data-corrected samples in the continuous samples. ^7. The data correction for increasing or decreasing data by an integral multiple of 2 ^MN-1 or less of 1/2 ^MN of the value of the least significant bit in the number N of effective bits. 9. The bit length extension device according to any one of claims 8 to 8. 10. The data correction unit according to claim 6, wherein, when the samples to be subjected to the data correction are consecutive, the data correction unit determines a value to be increased or decreased so that the data corrected does not have the same value. The bit length extension method according to any one of the above. 11. The data correction unit according to claim 1, wherein the value for increasing (or decreasing) the data of the sample for which data correction is to be performed is a value that is １ ／ of the value of the least significant bit in the effective bit number N, and The value to be increased (or increased) is の of the value of the least significant bit in the number N of effective bits.
^10. The bit length extension method according to claim 6, wherein the value is (2 ^MN-1 -1) times the value of ^MN .