JP3297792B2 - Signal expansion apparatus and method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION The present invention relates to, for example, a compact disc (CD) and a digital audio tape (DA).
The present invention relates to a signal decompression device and method capable of decompressing and outputting the word length of digital audio data recorded on a recording medium such as T).

[0002]

2. Description of the Related Art At present, recording media such as CDs and DATs include:
Data is recorded in a 16-bit recording format. In addition, professional recorders and AD converters, D
In the A converter, data is processed in 20 bits. For example, when converting digital audio data recorded by a recorder of 16 bits or more into CDs, simply discard the data of 16 bits or less or use data of 16 bits or less by super bit mapping (SBM) or the like. A method of compressing to bits is used.

By the way, from 16-bit data, 16 bits
There is no way to generate more data than bits. For example,
Many digital mixers and DA converters can process 20-bit or 24-bit data as described above. Even if such a device is supplied with data whose word tone has been truncated or compressed data, it can be processed only with the word length. Further, even if the data is simply quantized using an A / D converter according to the theoretical value, the data cannot be processed with a resolution greater than the word length of the A / D converter.

[0004]

Therefore, there is a method for solving the above-mentioned problem by using Lagrange interpolation, for example. In this method, two interpolation filters having the same characteristics and performing Lagrange interpolation are used. In the first interpolation filter, data at an intermediate point between sampling points of input data is interpolated with a longer word length than original data based on data of several samples before and after. Further, the second interpolation filter uses the interpolated data of several samples before and after at an intermediate point (sample point of the original signal) of the data interpolated by the first interpolation filter,
It is again interpolated to a word length longer than the word length of the original signal. Thereby, digital audio data having a word length longer than the data word length of the original signal can be obtained. Note that a sampling frequency (hereinafter, referred to as SF) used for the above-described interpolation is used.
Is 44.1 kHz, for example.

FIG. 4 is a characteristic diagram of output data when the word length is expanded using the above-described interpolation filter. In addition,
X axis is frequency (Hz), Y axis is output signal level (dB)
And SF are 44.1 kHz. As can be seen from FIG. 4, although the output signal level is constant below 15 kHz, the output signal level attenuates from the vicinity of over 15 kHz. This is due to the frequency characteristics of Lagrange interpolation.

FIG. 5 is a diagram showing frequency components of a signal sampled by the SF. Note that the X-axis is frequency and the Y-axis is signal level. As is clear from FIG. 5, this frequency component exists right before SF / 2. When the above-described Lagrange interpolation circuit is used for such a signal, a high-frequency component is attenuated.

It is therefore an object of the present invention to provide a decompression apparatus and method capable of decompressing the word length of data without attenuating high-frequency component levels even using Lagrange interpolation.

[0008]

SUMMARY OF THE INVENTION The present invention provides an oversampling filter 1 for increasing the sampling frequency of N-bit input data and a sampling point for the oversampled N-bit input data. DSP2 that performs interpolation of N-bit input data with long M bits and re-interpolates with L-bit data longer than N bits and shorter than M bits based on sampling points of the input data interpolated with M bits And a decimation filter for returning the sampling frequency of the output signal of the DSP 2 to the original sampling.

Further, the present invention increases the sampling frequency of N-bit input data and interpolates N-bit input data with M bits longer than the input data based on the sampling points of the N-bit input data. , M
On the basis of the sampling points of the input data interpolated by the bits, interpolation is performed again with L-bit data longer than N bits and shorter than M bits, and the sampling frequency of the interpolated signal is returned to the original sampling. This is the signal expansion method used.

[0010]

The SF of the input data X (n) is multiplied by N in the oversampling filter 1. DSP (Digi
tal Signal Processor) 2, SF = 1 / (N · S
F), data X (n) is sampled and data X (n) is sampled.
Data Y (n) in the middle of the sampling point (n)
Is generated. Further, data Y (n) is SF = 1 /
The sampling is performed at (N · SF), and data Z (n) is generated at an intermediate point (X (n)) of the sampling points. The data Z (n) is supplied to the decimation filter 3, and the sampling frequency set to N times is down-sampled to 1 / N times.

[0011]

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a block diagram showing a signal expansion apparatus according to the present invention. FIG. 1 is a block diagram of a signal decompression device according to the present invention. In FIG. 1, a predetermined format (for example, AES / EB) is input via an input terminal.
The digital audio data X (n) demodulated into the U format or the IEC958 format is supplied to the oversampling filter 1. In the oversampling filter 1, the SF of the data X (n) is N
Multiplied. Output data of the oversampling filter 1 is supplied to the DSP 2.

The DSP 2 comprises two Lagrange interpolation filters of the same kind. As the interpolation filter, for example, an FIR filter can be used. In the first interpolation filter, based on data of several samples before and after data X (n) (for example, 16 bits), data X
The data at the intermediate point of the sampling point (n) is interpolated with a longer word length than the data X (n). As a result, data Y (n) (for example, 48 bits) is generated. Further, in the second interpolation filter, at the intermediate point (the sampling point of the original signal) of the point interpolated by the first interpolation filter, the data Y (n) is again converted to the data X based on the data of several samples before and after it.
Data Z (n) having a word length longer than the word length of (n) (for example, 2
4 bits). Thereby, the data X (n)
Digital audio data having a word length longer than the word length can be obtained. Note that the SF used for the above interpolation is
Is, for example, SF = (N × 44.1) kHz.
Data Z (n) sampled at 1 / (N · SF)
Is supplied to the decimation filter 3.

After the SF multiplied by N in the oversampling filter 1 is down-sampled by 1 / N in the decimation filter 3, the data is supplied to the output terminal.

FIG. 2 is a diagram showing frequency components of a signal sampled by N times SF. Note that the X-axis is frequency and the Y-axis is signal level. By sampling at (N · SF), the frequency component is located on the lower frequency side with respect to SF. Considering the sampling frequency of (N · SF) as a center, the apparent frequency component of the signal is 1 / N. Therefore, even if the Lagrange interpolation in which the high-frequency level is attenuated is used, the reduction of the high-frequency level of the signal can be prevented without affecting the frequency component band of the signal at all.

Hereinafter, an interpolation method in the DSP 2 will be described with reference to FIG. The data X (n) output from the oversampling filter 1 is supplied to the DSP 2. As shown in FIG. 3A, data X indicated by a black circle
(N) is data sampled at a time of 1 / (N · SF). In the first Lagrange interpolation filter in the DPS2, the data at several points before and after the data X (n) is used to interpolate an intermediate point between the sampling points (see FIG. 3B). Thereby, data Y (n) indicated by a white circle is generated. In addition, the interpolation data Y (n) is obtained by performing interpolation calculation on M bits (N << M) longer than the word length of the data X (n), and then rounding to L bits (N <L <M) (or Truncation). Further, the word length of M bits is set to such a length that an error can be ignored for calculating data of L bits. For example, data Y (n) can be obtained by performing the above-described processing using four front points and three rear points with data X (n) as the center.

As an example, a generation formula when performing seventh-order Lagrange interpolation is shown below.

Y (n) = k0 × X (n−4) + k1 × X
(N−3) + k2 × X (n−2) + k3 × X (n−1)
+ K4 × X (n) + k5 × X (n + 1) + k6 × X (n
+2) + k7 × X (n + 3)

## EQU1 ## However,

K0 = -c1 * c2 * c3 * c4 * c5 * c6 * c7 / 35 k1 = c0 * c2 * c3 * c4 * c5 * c6 * c7 * 4/15 k2 = -c0 * c1 * c3 * c4 × c5 × c6 × c7 × 6/5 k3 = c0 × c1 × c2 × c4 × c5 × c6 × c7 × 4 k4 = c0 × c1 × c2 × c3 × c5 × c6 × c7 × 4 k5 = -c0 × c1 × c2 × c3 × c4 × c6 × c7 × 6/5 k6 = c0 × c1 × c2 × c3 × c4 × c5 × c7 × 4/15 k7 = -c0 × c1 × c2 × c3 × c4 × c5 × c6 / 35

C0 = (3 + 0.5) / 4 c1 = (2 + 0.5) / 3 c2 = (1 + 0.5) / 2 c3 = 0.5 c4 = 1-0.5 c5 = 1-0. 5/2 c6 = 1-0.5 / 3 c7 = 1-0.5 / 4

It is assumed that

Next, by interpolating at the intermediate point of the data Y (n) under the same conditions as described above, the data Z (n) shown in FIG. 3C can be obtained. Note that data Y
An intermediate point of (n), that is, a point of data Z (n) is the same as a sampling point of data X (n). By performing such interpolation, N-bit (for example, 16-bit) input data (X (n)) can be output as L-bit (for example, 24-bit) data having an expanded word length.

[0023]

According to the present invention, even when a Lagrange interpolation filter is used as a means for extending the word length of input data, the level in a high frequency range is not affected by the frequency characteristics. Interpolation can be performed without attenuation. Therefore, the S / N ratio (sound quality) can be improved.

[Brief description of the drawings]

FIG. 1 is a block diagram of a signal decompression device according to the present invention.

FIG. 2 is a diagram illustrating frequency components of signal components sampled by a signal decompression device.

FIG. 3 is a waveform diagram of input data and output data for a DSP.

FIG. 4 is a characteristic diagram of output data when a word length is extended using a conventional interpolation filter.

FIG. 5 is a diagram showing frequency components of signal components sampled by a conventional interpolation filter.

[Explanation of symbols]

 1 oversampling filter 2 DSP 3 decimation filter

Continuation of the front page (51) Int.Cl. ⁷ Identification symbol FI H03H 17/02 (56) References JP-A-64-27307 (JP, A) JP-A-2-117216 (JP, A) JP-A-2 -200004 (JP, A) JP-A-4-192711 (JP, A) JP-A-6-13991 (JP, A) JP-A-6-120776 (JP, A) JP-A-6-252701 (JP, A) (58) Field surveyed (Int.Cl. ⁷ , DB name) H03M 7/30 H03H 17/02

Claims (3)

(57) [Claims] An oversampling means for increasing a sampling frequency of N-bit input data; and M bits longer than the N-bit input data based on a sampling point of the oversampled N-bit input data. Interpolating means for interpolating the input data of the above, and interpolating again with L-bit data longer than the N-bits and shorter than the M-bits based on the sampling points of the input data interpolated by the M-bits; A signal decompression device comprising: downsampling means for returning the sampling frequency of the output signal of the interpolation means to the original sampling. 2. The signal decompression device according to claim 1, wherein said input data is digital audio data. 3. The sampling frequency of the N-bit input data is increased, and the N-bit input data is interpolated by M bits longer than the input data based on the sampling points of the N-bit input data. Based on the sampling points of the input data interpolated by the M bits, L longer than the N bits and shorter than the M bits
A signal expansion method in which interpolation is performed again using bit data, and the sampling frequency of the signal after the interpolation is returned to the original sampling.