JP6439843B2 - Signal processing apparatus and method, and program - Google Patents

Description

  The present invention relates to a signal processing apparatus, method, and program, and more particularly, to a signal processing apparatus, method, and program that can obtain higher-quality sound when an encoded audio signal is decoded.

  Conventionally, HE-AAC (High Efficiency MPEG (Moving Picture Experts Group) 4 AAC (Advanced Audio Coding)) (international standard ISO / IEC14496-3) or the like is known as an audio signal encoding method. In such an encoding method, a high-frequency feature encoding technique called SBR (Spectral Band Replication) is used (for example, see Patent Document 1).

  In SBR, at the time of encoding an audio signal, a low frequency component (hereinafter referred to as a low frequency signal) of the encoded audio signal and a high frequency component (hereinafter referred to as a high frequency signal) of the audio signal are generated. SBR information is output. The decoding device decodes the encoded low frequency signal, generates a high frequency signal using the low frequency signal obtained by decoding and the SBR information, and generates an audio signal composed of the low frequency signal and the high frequency signal. obtain.

  Specifically, for example, it is assumed that the low frequency signal SL1 shown in FIG. 1 is obtained by decoding. In FIG. 1, the horizontal axis indicates the frequency, and the vertical axis indicates the energy of each frequency of the audio signal. In the drawing, the dotted line in the vertical direction represents the boundary of the scale factor band. The scale factor band is a band obtained by bundling a plurality of subbands having a predetermined bandwidth, which is the resolution of a QMF (Quadrature Mirror Filter) analysis filter.

  In FIG. 1, in the figure of the low-frequency signal SL1, a band composed of seven continuous scale factor bands on the right side is defined as a high frequency, and by decoding SBR information, a high frequency is obtained for each scale factor band on the high frequency side. Scale factor band energies E11 to E17 are obtained.

  Then, the low frequency signal SL1 and the high frequency scale factor band energy are used to generate a high frequency signal of each scale factor band. For example, when a high-frequency signal of the scale factor band Bobj is generated, the component of the scale factor band Borg in the low-frequency signal SL1 is frequency-shifted to the band of the scale factor band Bobj, and a signal obtained by the frequency shift is obtained. The gain is adjusted to obtain a high frequency signal. At this time, gain adjustment is performed so that the average energy of the signal obtained by the frequency shift becomes the same as the high-frequency scale factor band energy E13 of the scale factor band Bobj.

  Through such processing, the high frequency signal SH1 shown in FIG. 2 is generated as a component of the scale factor band Bobj. In FIG. 2, the same reference numerals are given to the portions corresponding to those in FIG. 1, and the description thereof is omitted.

  In this way, on the decoding side of the audio signal, by using the low frequency signal and the SBR information, by generating a high frequency component not included in the encoded and decoded low frequency signal and extending the band, Higher quality sound can be played back.

JP-T-2001-521648

  However, when the low-frequency signal SL1 used for generating the high-frequency signal has a hole as in the scale factor band Borg in FIG. 2, the shape of the obtained high-frequency signal SH1 is the frequency shape of the original signal. There is a high possibility that the shapes will be greatly different, and this will cause deterioration in hearing. It should be noted that the state where the low-frequency signal has a hole is that the energy in the predetermined band is significantly smaller than the energy in the adjacent band, and a part of the low-frequency power spectrum (energy waveform of each frequency) In the figure, it is a state protruding downward, that is, a state where energy of some band components is recessed.

  In the example of FIG. 2, since the low-frequency signal SL1 used for generating the high-frequency signal has a dent, the dent also occurs in the high-frequency signal SH1. Thus, if there is a dent in the low-frequency signal used to generate the high-frequency signal, the high-frequency component cannot be accurately reproduced, and the audio signal obtained by decoding may be deteriorated in audibility.

  In SBR, processing called gain limiter or interpolation may be performed, and dents may occur in high frequency components due to these processing.

  Here, the gain limiter is a process of suppressing the gain peak value to the average value of the gain within the limiter band within the limiter band composed of a plurality of subbands.

  For example, it is assumed that the low frequency signal SL2 shown in FIG. 3 is obtained by decoding the low frequency signal. In FIG. 3, the horizontal axis indicates the frequency, and the vertical axis indicates the energy of each frequency of the audio signal. In the drawing, the dotted line in the vertical direction represents the boundary of the scale factor band.

  In FIG. 3, the band composed of seven consecutive scale factor bands on the right side of the low-frequency signal SL2 is a high frequency band, and high-frequency scale factor band energy E21 to E27 is obtained by decoding the SBR information. It is done.

  A band composed of the three scale factor bands Bobj1 to Bobj3 is a limiter band. Furthermore, it is assumed that the respective components of the scale factor bands Borg1 to Borg3 of the low frequency signal SL2 are used to generate the high frequency signals of the scale factor bands Bobj1 to Bobj3 on the high frequency side.

  Therefore, basically, when the high frequency signal SH2 of the scale factor band Bobj2 is generated, the gain adjustment is performed by the energy ratio G2 between the average energy of the scale factor band Borg2 of the low frequency signal SL2 and the high frequency scale factor band energy E22. Is done. That is, the component of the scale factor band Borg2 of the low-frequency signal SL2 is frequency-shifted, and the resulting signal is multiplied by the energy ratio G2 to perform gain adjustment to obtain the high-frequency signal SH2.

  However, in the gain limiter, when the energy ratio G2 is larger than the average value G of the energy ratios G1 to G3 of the scale factor bands Bobj1 to Bobj3 in the limiter band, the energy ratio G2 multiplied by the signal after the frequency shift is The average value G is assumed. That is, the gain of the high frequency signal of the scale factor band Bobj2 is kept low.

  In the example of FIG. 3, the energy of the scale factor band Borg2 of the low-frequency signal SL2 is smaller than the energy of the adjacent scale factor bands Borg1 and Borg3. In other words, the scale factor band Borg2 has a dent.

  On the other hand, the high frequency scale factor band energy E22 of the scale factor band Bobj2 to which the low frequency component is pasted is larger than the high frequency scale factor band energy of the scale factor bands Bobj1 and Bobj3.

  Therefore, the energy ratio G2 of the scale factor band Bobj2 becomes higher than the average value G of the energy ratios in the limiter band, and the gain of the high frequency signal of the scale factor band Bobj2 is suppressed to a low level by the gain limiter.

  Therefore, in the scale factor band Bobj2, the energy of the high frequency signal SH2 is significantly lower than the high frequency scale factor band energy E22, and the frequency shape of the generated high frequency signal is significantly different from the frequency shape of the original signal. It becomes a shape. If it does so, the audio | voice finally obtained by decoding will produce deterioration on hearing.

  Interpolation is a high-frequency signal generation method in which frequency shift and gain adjustment are performed for each subband, not for each scale factor band.

  For example, as shown in FIG. 4, each of the subbands Borg1 to Borg3 of the low frequency signal SL3 is used to generate the high frequency signals of the high frequency subbands Bobj1 to Bobj3, and the subbands Bobj1 to Bobj3 are generated. Suppose that the band consisting of is a limiter band.

  In FIG. 4, the horizontal axis indicates the frequency, and the vertical axis indicates the energy of each frequency of the audio signal. Further, high frequency scale factor band energies E31 to E37 are obtained for each scale factor band by decoding of the SBR information.

  In the example of FIG. 4, the energy of the subband Borg2 of the low-frequency signal SL3 is smaller than the energy of the adjacent subbands Borg1 and Borg3, and a dent is generated in the subband Borg2. Therefore, as in the case of FIG. 3, the energy ratio between the energy of the sub-band Borg2 of the low-frequency signal SL3 and the high-frequency scale factor band energy E33 is higher than the average value of the energy ratio in the limiter band. Then, the gain of the high-frequency signal SH3 of the subband Bobj2 is suppressed by the gain limiter.

  As a result, in the subband Bobj2, the energy of the high frequency signal SH3 is significantly lower than the high frequency scale factor band energy E33, and the frequency shape of the generated high frequency signal is significantly different from the frequency shape of the original signal. It can be a shape. As a result, as in the case of FIG. 3, the audio obtained by decoding is deteriorated in terms of hearing.

  As described above, in the SBR, depending on the shape (frequency shape) of the power spectrum of the low-frequency signal used for generating the high-frequency signal, high-quality sound may not be obtained on the audio signal decoding side.

  The present invention has been made in view of such a situation, and is intended to obtain higher-quality sound when an audio signal is decoded.

  A signal processing device according to one aspect of the present technology includes an extraction unit that extracts a low frequency component of an audio signal and high frequency information for obtaining the high frequency component of the audio signal, and QMF analysis filter processing on the low frequency component And a flattening processing unit for flattening the low-frequency subband signal based on energy and average energy of the low-frequency subband signal, Further, based on the frequency shift unit that frequency-shifts the low-frequency subband signal to the high frequency, the signal frequency-shifted to the high frequency and the high-frequency information, the gain of the signal frequency-shifted to the high frequency is A high-frequency generating unit that generates a high-frequency sub-band signal by adjusting, and a QMF synthesis filter processing unit that generates an audio signal by synthesizing the low-frequency sub-band signal and the high-frequency sub-band signal by a QMF synthesis filter And Obtain.

  A signal processing method or program according to one aspect of the present technology extracts a low frequency component of an audio signal and high frequency information for obtaining the high frequency component of the audio signal, and performs QMF analysis filtering on the low frequency component. To generate a low-frequency subband signal, flatten the low-frequency subband signal based on the energy and average energy of the low-frequency subband signal, and frequency-shift the flattened low-frequency subband signal to a high frequency. And generating a high frequency sub-band signal by adjusting a gain of the signal frequency shifted to the high frequency based on the signal frequency-shifted to the high frequency and the high frequency information, and generating the low frequency sub-band signal. And a step of synthesizing the signal and the high frequency sub-band signal by a QMF synthesis filter to generate an audio signal.

  In one aspect of the present technology, a low-frequency component of an audio signal and high-frequency information for obtaining the high-frequency component of the audio signal are extracted, and the low-frequency component is subjected to QMF analysis filter processing to obtain a low-frequency subband. A signal is generated, the low frequency subband signal is flattened based on the energy and average energy of the low frequency subband signal, the flattened low frequency subband signal is frequency shifted to a high frequency, and the high frequency subband signal is Based on the frequency-shifted signal and the high-frequency information, the gain of the signal frequency-shifted to the high frequency is adjusted to generate a high-frequency subband signal, the low-frequency subband signal, The high frequency sub-band signal is synthesized by the QMF synthesis filter to generate an audio signal.

  According to one aspect of the present invention, it is possible to obtain higher-quality sound when decoding an audio signal.

It is a figure explaining conventional SBR. It is a figure explaining conventional SBR. It is a figure explaining the conventional gain limiter. It is a figure explaining the conventional interpolation. It is a figure explaining SBR to which the present invention is applied. It is a figure which shows the structural example of one Embodiment of the encoder to which this invention is applied. It is a flowchart explaining an encoding process. It is a figure which shows the structural example of one Embodiment of the decoder to which this invention is applied. It is a flowchart explaining a decoding process. It is a flowchart explaining an encoding process. It is a flowchart explaining a decoding process. It is a flowchart explaining an encoding process. It is a flowchart explaining a decoding process. It is a block diagram which shows the structural example of a computer.

  Embodiments to which the present invention is applied will be described below with reference to the drawings.

<Outline of the present invention>
First, with reference to FIG. 5, description will be given of band expansion of an audio signal by SBR to which the present invention is applied. In FIG. 5, the horizontal axis indicates the frequency, and the vertical axis indicates the energy of each frequency of the audio signal. In the drawing, the dotted line in the vertical direction represents the boundary of the scale factor band.

  For example, on the audio signal decoding side, the low frequency signal SL11 of the audio signal and the high frequency scale factor band energies Eobj1 to Eobj7 of the scale factor bands Bobj1 to Bobj7 on the high frequency side are obtained from the data received from the encoding side. Suppose that Then, it is assumed that the low frequency signal SL11 and the high frequency scale factor band energy Eobj1 to Eobj7 are used to generate high frequency signals of the respective scale factor bands Bobj1 to Bobj7.

  Now, let us consider generating a high frequency signal of the scale factor band Bobj3 on the high frequency side using the components of the scale factor band Borg1 of the low frequency signal SL11.

  In the example of FIG. 5, the power spectrum of the low-frequency signal SL11 is greatly dented downward in the figure in the scale factor band Borg1 portion. That is, the energy is small compared to other bands. For this reason, when the high frequency signal of the scale factor band Bobj3 is generated by the conventional SBR, the resulting high frequency signal is also dented, resulting in a deterioration in the audibility of the sound.

  Therefore, in the present invention, first, flattening processing (smoothing processing) is performed on the components of the scale factor band Borg1 of the low-frequency signal SL11. As a result, the low-frequency signal H11 of the scale factor band Borg1 after flattening is obtained. The power spectrum of the low frequency signal H11 is smoothly connected to the portion of the band adjacent to the scale factor band Borg1 in the power spectrum of the low frequency signal SL11. That is, the flattened low-frequency signal SL11 has no dent in the scale factor band Borg1.

  In this way, when the low frequency signal SL11 is flattened, the low frequency signal H11 obtained by the flattening is frequency shifted to the band of the scale factor band Bobj3, and the signal obtained by the frequency shift is gain adjusted. Thus, a high frequency signal H12 is obtained.

  At this time, the average value of the energy of each subband of the low frequency signal H11 is obtained as the average energy Eorg1 of the scale factor band Borg1. Then, the gain of the low frequency signal H11 after the frequency shift is adjusted according to the ratio between the average energy Eorg1 and the high frequency scale factor band energy Eobj3. Specifically, the gain adjustment is performed so that the average value of the energy of each subband of the frequency-shifted low frequency signal H11 becomes substantially the same as the high frequency scale factor band energy Eobj3.

  In FIG. 5, since the high frequency signal H12 is generated by using the low frequency signal H11 having no dent, the energy of each subband of the high frequency signal H12 is substantially the same as the high frequency scale factor band energy Eobj3. It has become. Therefore, a high frequency signal substantially the same as the high frequency signal of the original signal is obtained.

  Thus, if the high frequency signal is generated using the flattened low frequency signal, the high frequency component of the audio signal can be generated with higher accuracy. Conventionally, the low frequency signal has a dent in the power spectrum. It is possible to improve the audible degradation of the generated audio signal. That is, higher quality sound can be obtained.

  In addition, if the low frequency signal is flattened, dents in the power spectrum can be removed. Therefore, if a high frequency signal is generated using the flattened low frequency signal, a gain limiter or interpolation is performed. Even in this case, it is possible to prevent auditory degradation of the audio signal.

  Note that the flattening of the low-frequency signal may be performed on the entire low-frequency band component used for generating the high-frequency signal, or the band in which the dent is generated in the low-frequency band component. It may be performed only on the component. In addition, when flattening is performed only on the band component in which the dent is generated, the band to be flattened may be one subband or a plurality of subbands as long as the band is a subband unit. It may be a band having an arbitrary width.

  Further, hereinafter, regarding a band composed of several subbands such as a scale factor band, an average value of energy of each subband constituting the band is also referred to as an average energy of the band.

  Next, an encoder and a decoder to which the present invention is applied will be described. In the following description, a case where high-frequency signals are generated in units of scale factor bands will be described as an example. However, high-frequency signals are generated for each band composed of one or a plurality of subbands. Is possible.

<First Embodiment>
[Configuration of encoder]
FIG. 6 is a diagram showing a configuration example of an embodiment of an encoder to which the present invention is applied.

  The encoder 11 includes a downsampler 21, a low frequency encoding circuit 22, a QMF analysis filter processing unit 23, a high frequency encoding circuit 24, and a multiplexing circuit 25. An input signal that is an audio signal is supplied to the downsampler 21 and the QMF analysis filter processing unit 23 of the encoder 11.

  The downsampler 21 extracts a low-frequency signal that is a low-frequency component of the input signal by down-sampling the supplied input signal and supplies the low-frequency signal to the low-frequency encoding circuit 22. The low frequency encoding circuit 22 encodes the low frequency signal supplied from the down sampler 21 by a predetermined encoding method, and supplies the low frequency encoded data obtained as a result to the multiplexing circuit 25. As a method of encoding the low frequency signal, for example, there is an AAC method.

  The QMF analysis filter processing unit 23 performs a filter process using a QMF analysis filter on the supplied input signal, and divides the input signal into a plurality of subband signals. For example, the entire frequency band of the input signal is divided into 64 by filtering, and the components of those 64 bands (subbands) are extracted. The QMF analysis filter processing unit 23 supplies the signal of each subband obtained by the filter processing to the high frequency encoding circuit 24.

  Hereinafter, each subband signal of the input signal is also referred to as a subband signal. In particular, the band of the low frequency signal extracted by the down sampler 21 is defined as a low frequency, and the subband signal of each subband on the low frequency side is referred to as a low frequency subband signal. Further, of the entire band of the input signal, a band having a frequency higher than that of the low frequency band is defined as a high frequency, and the subband signal of the high frequency subband is referred to as a high frequency subband signal.

  Furthermore, in the following, the description will be continued with a band having a higher frequency than the low band as a high band, but a part of the low band and the high band may overlap. That is, the low frequency band and the high frequency band may include a common band.

  The high frequency encoding circuit 24 generates SBR information based on the subband signal supplied from the QMF analysis filter processing unit 23 and supplies the SBR information to the multiplexing circuit 25. Here, the SBR information is information for obtaining the high frequency scale factor band energy of each scale factor band on the high frequency side of the input signal that is the original signal.

  The multiplexing circuit 25 multiplexes the low frequency encoded data from the low frequency encoding circuit 22 and the SBR information from the high frequency encoding circuit 24 and outputs a bit stream obtained by multiplexing.

[Description of encoding process]
By the way, when an input signal is input to the encoder 11 and the encoding of the input signal is instructed, the encoder 11 performs an encoding process to encode the input signal. Hereinafter, the encoding process by the encoder 11 will be described with reference to the flowchart of FIG.

  In step S <b> 11, the downsampler 21 extracts the low frequency signal by down-sampling the supplied input signal and supplies the low frequency signal to the low frequency encoding circuit 22.

  In step S <b> 12, the low frequency encoding circuit 22 encodes the low frequency signal supplied from the down sampler 21 by, for example, the AAC method, and supplies the low frequency encoded data obtained as a result to the multiplexing circuit 25.

  In step S <b> 13, the QMF analysis filter processing unit 23 performs a filter process using the QMF analysis filter on the supplied input signal, and outputs the subband signal of each subband obtained as a result thereof to the high frequency encoding circuit 24. To supply.

  In step S14, the high frequency encoding circuit 24 obtains a high frequency scale factor band energy Eobj of each scale factor band on the high frequency side based on the subband signal supplied from the QMF analysis filter processing unit 23.

  That is, the high frequency encoding circuit 24 uses a band composed of several continuous subbands on the high frequency side as a scale factor band, and uses a subband signal of each subband in the scale factor band to Calculate energy. Then, the high frequency encoding circuit 24 calculates the average value of the energy of each subband in the scale factor band, and sets the calculated average value of the energy as the high frequency scale factor band energy Eobj of the scale factor band. Thereby, for example, the high frequency scale factor band energy Eobj1 to Eobj7 of FIG. 5 is calculated.

  In step S15, the high frequency encoding circuit 24 encodes the high frequency scale factor band energy Eobj of a plurality of scale factor bands by a predetermined encoding method to generate SBR information. For example, the high-frequency scale factor band energy Eobj is encoded by a method such as scalar quantization, differential encoding, or variable length encoding. The high frequency encoding circuit 24 supplies SBR information obtained by encoding to the multiplexing circuit 25.

  In step S16, the multiplexing circuit 25 multiplexes the low frequency encoded data from the low frequency encoding circuit 22 and the SBR information from the high frequency encoding circuit 24, and the bit stream obtained by multiplexing is multiplexed. Output, and the encoding process ends.

  In this way, the encoder 11 encodes the input signal and outputs a bit stream in which the low frequency encoded data and the SBR information are multiplexed. Therefore, on the receiving side of this bit stream, the low frequency encoded data is decoded to obtain a low frequency signal, and a high frequency signal is generated using the low frequency signal and the SBR information. It is possible to obtain a wider-band audio signal composed of signals.

[Decoder configuration]
Next, a decoder that receives and decodes the bitstream output from the encoder 11 of FIG. 6 will be described. For example, the decoder is configured as shown in FIG.

  That is, the decoder 51 includes a demultiplexing circuit 61, a low frequency decoding circuit 62, a QMF analysis filter processing unit 63, a high frequency decoding circuit 64, and a QMF synthesis filter processing unit 65.

  The demultiplexing circuit 61 demultiplexes the bit stream received from the encoder 11 and extracts low frequency encoded data and SBR information. The demultiplexing circuit 61 supplies the low frequency encoded data obtained by the demultiplexing to the low frequency decoding circuit 62 and supplies the SBR information to the high frequency decoding circuit 64.

  The low frequency decoding circuit 62 decodes the low frequency encoded data supplied from the non-multiplexing circuit 61 by a decoding method corresponding to the low frequency signal encoding method (for example, AAC method) used in the encoder 11, The low frequency signal obtained as a result is supplied to the QMF analysis filter processing unit 63. The QMF analysis filter processing unit 63 performs filter processing using a QMF analysis filter on the low frequency signal supplied from the low frequency decoding circuit 62, and subband signals of each subband from the low frequency signal to the low frequency side. To extract. That is, band division of the low frequency signal is performed. The QMF analysis filter processing unit 63 supplies the low-frequency subband signal of each subband on the low frequency side obtained by the filter processing to the high-frequency decoding circuit 64 and the QMF synthesis filter processing unit 65.

  The high frequency decoding circuit 64 uses the SBR information supplied from the non-multiplexing circuit 61 and the low frequency sub-band signal supplied from the QMF analysis filter processing unit 63 to increase the high frequency of each scale factor band on the high frequency side. A band signal is generated and supplied to the QMF synthesis filter processing unit 65.

  The QMF synthesis filter processing unit 65 synthesizes the low frequency sub-band signal supplied from the QMF analysis filter processing unit 63 and the high frequency signal supplied from the high frequency decoding circuit 64 by filter processing using a QMF synthesis filter. To generate an output signal. This output signal is an audio signal composed of low-frequency and high-frequency subband components, and the output signal is output from the QMF synthesis filter processing unit 65 to a reproduction unit such as a speaker at the subsequent stage.

[Description of decryption processing]
When a bit stream is supplied from the encoder 11 to the decoder 51 shown in FIG. 8 and decoding of the bit stream is instructed, the decoder 51 performs a decoding process to generate an output signal. Hereinafter, the decoding process by the decoder 51 will be described with reference to the flowchart of FIG.

  In step S41, the demultiplexing circuit 61 demultiplexes the bit stream received from the encoder 11. Then, the demultiplexing circuit 61 supplies the low frequency encoded data obtained by demultiplexing the bitstream to the low frequency decoding circuit 62 and also supplies the SBR information to the high frequency decoding circuit 64.

  In step S 42, the low frequency decoding circuit 62 decodes the low frequency encoded data supplied from the demultiplexing circuit 61 and supplies the low frequency signal obtained as a result to the QMF analysis filter processing unit 63.

  In step S43, the QMF analysis filter processing unit 63 performs a filter process using the QMF analysis filter on the low frequency signal supplied from the low frequency decoding circuit 62. Then, the QMF analysis filter processing unit 63 supplies the low frequency subband signal of each subband on the low frequency side obtained as a result of the filter processing to the high frequency decoding circuit 64 and the QMF synthesis filter processing unit 65.

  In step S44, the high frequency decoding circuit 64 decodes the SBR information supplied from the non-multiplexing circuit 61. Thereby, the high frequency scale factor band energy Eobj of each scale factor band on the high frequency side is obtained.

  In step S45, the high frequency decoding circuit 64 performs a flattening process on the low frequency sub-band signal supplied from the QMF analysis filter processing unit 63.

  For example, for the high-frequency side scale factor band, the high-frequency decoding circuit 64 converts the low-frequency side scale factor band used for generating the high-frequency signal of the scale factor band into the target scale factor for flattening processing. A band. It is assumed that the scale factor band on the low frequency side used for generating the high frequency signal of each scale factor band on the high frequency side is predetermined.

  Next, the high frequency decoding circuit 64 performs a filtering process using a flattening filter on the low frequency sub-band signals of each sub-band constituting the scale factor band to be processed on the low frequency side. Specifically, the high frequency decoding circuit 64 obtains the energy of each subband based on the low frequency subband signal of each subband constituting the scale factor band to be processed on the low frequency side. The average value of the energy of each subband is obtained as the average energy. The high frequency decoding circuit 64 multiplies the low frequency subband signal of each subband constituting the scale factor band to be processed by the ratio of the energy of these subbands to the average energy, thereby obtaining the low frequency of each subband. Flatten the subband signal.

  For example, it is assumed that the scale factor band to be processed includes three subbands SB1 to SB3, and energy E1 to E3 is obtained as energy of these subbands. In this case, the average value of the energy E1 to E3 of the subbands SB1 to SB3 is obtained as the average energy EA.

  Then, each of the low-frequency subband signals of subbands SB1 to SB3 is multiplied by EA / E1, EA / E2, and EA / E3, which are energy ratio values. In this way, the low-frequency subband signal multiplied by the energy ratio is a flattened low-frequency subband signal.

  Note that the low band subband signal may be flattened by multiplying the low band subband signal of the subband by the ratio of the maximum value of the energy E1 to E3 and the energy of the subband. The flattening of the low-frequency subband signal of each subband may be performed in any way as long as the power spectrum of the scale factor band composed of these subbands is flattened.

  In this way, for each high-frequency side scale factor band to be generated, the low-frequency sub-band signal of each sub-band constituting the low-frequency side scale factor band used for generating the scale factor band is Flattened.

  In step S46, the high frequency decoding circuit 64 obtains the average energy Eorg of the scale factor bands for each of the low frequency side scale factor bands used for generating the high frequency side scale factor band.

  Specifically, the high frequency decoding circuit 64 obtains the energy of each subband by using the low frequency subband signal after the flattening of each subband constituting the scale factor band on the low frequency side, The average value of the energy of the subband is obtained as the average energy Eorg.

  In step S47, the high frequency decoding circuit 64 uses the low frequency side scale factor band signal used for generating the high frequency side scale factor band to the frequency band of the high frequency side scale factor band to be generated. shift. That is, the low-frequency sub-band signals of the respective sub-bands that form the low-frequency scale factor band are frequency-shifted.

  In step S48, the high frequency decoding circuit 64 adjusts the gain of the low frequency sub-band signal after the frequency shift in accordance with the ratio of the high frequency scale factor band energy Eobj and the average energy Eorg, and adjusts the high frequency side scale factor band. Generate a high frequency sub-band signal.

  For example, a high-frequency scale factor band to be generated is called a high-frequency scale factor band, and a low-frequency scale factor band used to generate the high-frequency scale factor band is called a low-frequency scale factor band. I will do it.

  In the high frequency decoding circuit 64, the average value of the energy of the low frequency sub-band signal of each sub-band after the frequency shift constituting the low frequency scale factor band is approximately equal to the high frequency scale factor band energy of the high frequency scale factor band. The gain of the low-frequency subband signal after the frequency shift is adjusted so as to have the same magnitude.

  The low frequency sub-band signal frequency-shifted and gain-adjusted in this way becomes the high frequency sub-band signal of each sub-band of the high frequency scale factor band, and the high frequency of each sub-band of the scale factor band on the high frequency side. A signal composed of subband signals is a high-frequency scale factor band signal (high-frequency signal). The high frequency decoding circuit 64 supplies the generated high frequency signal of each scale factor band on the high frequency side to the QMF synthesis filter processing unit 65.

  In step S49, the QMF synthesis filter processing unit 65 uses the QMF synthesis filter for the low frequency sub-band signal supplied from the QMF analysis filter processing unit 63 and the high frequency signal supplied from the high frequency decoding circuit 64. The output signal is generated by combining by filtering. Then, the QMF synthesis filter processing unit 65 outputs the generated output signal, and the decoding process ends.

  In this way, the decoder 51 flattens the low-frequency subband signal, and generates a high-frequency signal for each scale factor band on the high-frequency side using the flattened low-frequency subband signal and the SBR information. . In this way, by generating a high frequency signal using the flattened low frequency sub-band signal, an output signal capable of reproducing higher quality sound can be easily obtained.

  In the above description, it has been described that the entire band on the low frequency side is flattened. However, on the decoder 51 side, the flattening may be performed only on the band where the dent is generated in the low frequency band. . In such a case, for example, the decoder 51 uses the low frequency signal to detect the frequency band in which the dent is generated.

<Second Embodiment>
[Description of encoding process]
In addition, the encoder 11 may generate position information of a band in which a dent is generated in the low band and information used for flattening the band, and output SBR information including the information. In such a case, the encoder 11 performs the encoding process shown in FIG.

  Hereinafter, with reference to the flowchart of FIG. 10, a description will be given of an encoding process in a case where SBR information including position information of a band in which a dent has occurred is output.

  In addition, since the process of step S71 thru | or step S73 is the same as the process of FIG.7 S11 thru | or step S13, the description is abbreviate | omitted. When the process of step S73 is performed, the subband signal of each subband is supplied to the high frequency encoding circuit 24.

  In step S74, the high frequency encoding circuit 24 uses the low frequency subband signal of the low frequency side subband supplied from the QMF analysis filter processing unit 23 to generate a dent band in the low frequency band. Is detected.

  Specifically, for example, the high frequency encoding circuit 24 determines the average energy EL that is the average value of the energy of the entire low frequency band by determining the average value of the energy of each subband of the low frequency band. Then, the high frequency encoding circuit 24 detects a subband in which the difference between the average energy EL and the energy of the subband is equal to or greater than a predetermined threshold among the low frequency subbands. That is, a subband whose value obtained by subtracting the energy of the subband from the average energy EL is equal to or greater than a threshold value is detected.

  Further, the high frequency encoding circuit 24 is a band composed of subbands in which the above-described difference is equal to or greater than a threshold, and a band composed of several consecutive subbands is converted into a band with depressions (hereinafter referred to as a flattened band). Called). The flattening band may be a band composed of one subband.

  In step S75, for each flattening band, the high-frequency encoding circuit 24 obtains flattening position information indicating the position of the flattening band and flattening gain information used for flattening the flattening band, and each flattening band. Information consisting of the flattening position information and the flattening gain information is used as flattening information.

  Specifically, the high frequency encoding circuit 24 uses the information indicating the flattened band as the flattened position information. Further, the high frequency encoding circuit 24 calculates the difference ΔE between the average energy EL and the energy of the subband for each subband constituting the flattened band, and the difference between the subbands constituting the flattened band. Information consisting of ΔE is set as flattening gain information.

  In step S76, the high frequency encoding circuit 24 obtains the high frequency scale factor band energy Eobj of each scale factor band on the high frequency side based on the subband signal supplied from the QMF analysis filter processing unit 23. In step S76, processing similar to that in step S14 in FIG. 7 is performed.

  In step S77, the high frequency encoding circuit 24 encodes the high frequency scale factor band energy Eobj of each scale factor band on the high frequency side and the flattening information of each flattened band by an encoding method such as scalar quantization. To generate SBR information. The high frequency encoding circuit 24 supplies the generated SBR information to the multiplexing circuit 25.

  Thereafter, the process of step S78 is performed, and the encoding process ends. However, the process of step S78 is the same as the process of step S16 in FIG.

  In this way, the encoder 11 detects the flattened band from the low band, and outputs SBR information including the flattened information used for flattening each flattened band together with the low band encoded data. As a result, the flattening band can be flattened more easily on the decoder 51 side.

[Description of decryption processing]
When the bit stream output by the encoding process described with reference to the flowchart of FIG. 10 is transmitted to the decoder 51, the decoder 51 that has received the bit stream performs the decoding process shown in FIG. Hereinafter, the decoding process by the decoder 51 will be described with reference to the flowchart of FIG.

  Note that the processing from step S101 to step S104 is the same as the processing from step S41 to step S44 in FIG. However, in the process of step S104, the high frequency scale factor band energy Eobj and the flattening information of each flattening band are obtained by decoding the SBR information.

  In step S105, the high frequency decoding circuit 64 uses the flattening information to flatten the flattened band indicated by the flattening position information included in the flattening information. That is, the high-frequency decoding circuit 64 performs flattening by adding the subband difference ΔE to the low-band subband signal of the subband constituting the flattened band indicated by the flattened position information. Here, the difference ΔE for each subband of the flattening band is information included as flattening gain information in the flattening information.

  Thus, when the low-frequency subband signal of each subband constituting the flattened band is flattened among the subbands on the low-frequency side, the flattened low-frequency subband signal is used thereafter. Thus, the process from step S106 to step S109 is performed, and the decoding process ends. Note that the processing from step S106 to step S109 is the same as the processing from step S46 to step S49 in FIG.

  In this manner, the decoder 51 performs flattening of the flattened band using the flattened information included in the SBR information, and generates a high frequency signal of each scale factor band on the high frequency side. As described above, by flattening the flattening band using the flattening information, a high frequency signal can be generated more easily and quickly.

<Third Embodiment>
[Description of encoding process]
Further, in the second embodiment, it has been described that the flattening information is included in the SBR information as it is and transmitted to the decoder 51. However, the flattening information may be vector-quantized and included in the SBR information.

  In such a case, for example, the high frequency encoding circuit 24 of the encoder 11 records a position table in which a plurality of flattened position information vectors are associated with position indexes that specify the flattened position information vectors. doing. Here, the flattened position information vector is a vector having each of the flattened position information of one or a plurality of flattened bands as an element, and the flattened position information is arranged in order from the lowest flattened band frequency. This is the resulting vector.

  In the position table, not only different flattened position information vectors composed of the same number of elements but also a plurality of flattened position information vectors composed of different numbers of elements are recorded.

  Further, the high frequency encoding circuit 24 of the encoder 11 records a gain table in which a plurality of flattening gain information vectors and gain indexes for specifying these flattening gain information vectors are associated with each other. Note that the flattening gain information vector is a vector having the flattening gain information of one or a plurality of flattening bands as elements, and is obtained by arranging the flattening gain information in ascending order of the flattening band frequencies. Vector.

  As in the case of the position table, a plurality of different flattening gain information vectors composed of the same number of elements and a plurality of flattening gain information vectors composed of different numbers of elements are also recorded in the gain table.

  Thus, when the position table and the gain table are recorded in the encoder 11, the encoder 11 performs the encoding process shown in FIG. Hereinafter, the encoding process by the encoder 11 will be described with reference to the flowchart of FIG.

  Note that the processes in steps S141 through S145 are the same as those in steps S71 through S75 in FIG.

  When the process of step S145 is performed, the flattened position information and the flattened gain information are obtained for each flattened band in the low band of the input signal. Then, the high frequency encoding circuit 24 arranges the flattened position information of each flattened band in order from the lowest frequency band to obtain a flattened position information vector, and the flattened gain of each flattened band in the order from the lowest frequency band The information is arranged into a flattening gain information vector.

  In step S146, the high frequency encoding circuit 24 acquires a position index and a gain index corresponding to the obtained flattened position information vector and flattened gain information vector.

  That is, the high frequency encoding circuit 24 selects the flattened position information vector having the shortest Euclidean distance from the flattened position information vector obtained in step S145 from the flattened position information vectors recorded in the position table. Identify. Then, the high frequency encoding circuit 24 acquires a position index associated with the specified flattened position information vector from the position table.

  Similarly, the high frequency encoding circuit 24 uses the flattening gain information vector having the shortest Euclidean distance from the flattening gain information vector obtained in step S145 from the flattening gain information vectors recorded in the gain table. Is identified. Then, the high frequency encoding circuit 24 acquires a gain index associated with the specified flattening gain information vector from the gain table.

  When the position index and the gain index are acquired in this way, the process of step S147 is subsequently performed to calculate the high frequency scale factor band energy Eobj of each scale factor band on the high frequency side. Note that the processing in step S147 is the same as the processing in step S76 in FIG.

  In step S148, the high-frequency encoding circuit 24 encodes each high-frequency scale factor band energy Eobj and the position index and gain index acquired in step S146 by an encoding method such as scalar quantization, and SBR information. Is generated. The high frequency encoding circuit 24 supplies the generated SBR information to the multiplexing circuit 25.

  Thereafter, the process of step S149 is performed, and the encoding process ends. However, the process of step S149 is the same as the process of step S78 of FIG.

  In this way, the encoder 11 detects the flattening band from the low band, and converts the SBR information including the position index and gain index to obtain the flattening information used for flattening each flattened band into the low band code. Output together with digitized data. Thereby, the information amount of the bit stream output from the encoder 11 can be reduced.

[Description of decryption processing]
Further, when the SBR information includes a position index and a gain index, a position table and a gain table are recorded in advance in the high frequency decoding circuit 64 of the decoder 51.

  As described above, when the decoder 51 records the position table and the gain table, the decoder 51 performs the decoding process shown in FIG. Hereinafter, the decoding process by the decoder 51 will be described with reference to the flowchart of FIG.

  Note that the processing from step S171 to step S174 is the same as the processing from step S101 to step S104 in FIG. However, in the process of step S174, the high frequency scale factor band energy Eobj, the position index, and the gain index are obtained by decoding the SBR information.

  In step S175, the high frequency decoding circuit 64 acquires a flattened position information vector and a flattened gain information vector based on the position index and the gain index.

  That is, the high frequency decoding circuit 64 acquires the flattened position information vector associated with the position index obtained by decoding from the recorded position table, and the gain index obtained by decoding from the gain table. The flattening gain information vector associated with is acquired. From the flattening position information vector and the flattening gain information vector thus obtained, flattening information of each flattening band, that is, flattening position information and flattening gain information of each flattening band is obtained.

  When the flattening information of each flattened band is obtained, the processing from step S176 to step S180 is performed thereafter, and the decoding process ends. These processes are the same as the processes from step S105 to step S109 in FIG. Therefore, the description thereof is omitted.

  In this way, the decoder 51 obtains the flattening information of each flattened band from the position index and gain index included in the SBR information, flattenes the flattened band, and each scale factor band on the high frequency side. Generate a high frequency signal. Thus, by obtaining the flattening information from the position index and the gain index, it is possible to reduce the information amount of the received bit stream.

  The series of processes described above can be executed by hardware or can be executed by software. When a series of processing is executed by software, a program constituting the software may execute various functions by installing a computer incorporated in dedicated hardware or various programs. For example, it is installed from a program recording medium in a general-purpose personal computer or the like.

  FIG. 14 is a block diagram illustrating a hardware configuration example of a computer that executes the above-described series of processing by a program.

  In a computer, a central processing unit (CPU) 201, a read only memory (ROM) 202, and a random access memory (RAM) 203 are connected to each other by a bus 204.

  An input / output interface 205 is further connected to the bus 204. The input / output interface 205 includes an input unit 206 including a keyboard, a mouse, and a microphone, an output unit 207 including a display and a speaker, a recording unit 208 including a hard disk and nonvolatile memory, and a communication unit 209 including a network interface. A drive 210 for driving a removable medium 211 such as a magnetic disk, an optical disk, a magneto-optical disk, or a semiconductor memory is connected.

  In the computer configured as described above, the CPU 201 loads, for example, the program recorded in the recording unit 208 to the RAM 203 via the input / output interface 205 and the bus 204, and executes the program. Is performed.

  The program executed by the computer (CPU 201) is, for example, a magnetic disk (including a flexible disk), an optical disk (CD-ROM (Compact Disc-Read Only Memory), DVD (Digital Versatile Disc), etc.), a magneto-optical disk, or a semiconductor. The program is recorded on a removable medium 211 that is a package medium composed of a memory or the like, or provided via a wired or wireless transmission medium such as a local area network, the Internet, or digital satellite broadcasting.

  The program can be installed in the recording unit 208 via the input / output interface 205 by attaching the removable medium 211 to the drive 210. Further, the program can be received by the communication unit 209 via a wired or wireless transmission medium and installed in the recording unit 208. In addition, the program can be installed in the ROM 202 or the recording unit 208 in advance.

  The program executed by the computer may be a program that is processed in time series in the order described in this specification, or in parallel or at a necessary timing such as when a call is made. It may be a program for processing.

  The embodiment of the present invention is not limited to the above-described embodiment, and various modifications can be made without departing from the gist of the present invention.

  11 Encoder, 22 Low frequency encoding circuit, 24 High frequency encoding circuit, 25 Multiplexing circuit, 51 Decoder, 61 Demultiplexing circuit, 63 QMF analysis filter processing unit, 64 High frequency decoding circuit, 65 QMF synthesis filter processing unit

Claims (3)

An extraction unit that extracts a low frequency component of the audio signal and high frequency information for obtaining a high frequency component of the audio signal;
A QMF analysis filter processing unit that generates a low-frequency subband signal by subjecting the low-frequency component to QMF analysis filter processing;
A flattening processing unit for flattening the low-frequency subband signal based on energy and average energy of the low-frequency subband signal;
A frequency shift unit that frequency-shifts the flattened low-frequency subband signal to a high frequency;
A high-frequency generation unit that generates a high-frequency sub-band signal by adjusting a gain of the signal frequency-shifted to the high frequency based on the signal frequency-shifted to the high frequency and the high frequency information;
A signal processing apparatus comprising: a QMF synthesis filter processing unit that synthesizes the low frequency subband signal and the high frequency subband signal with a QMF synthesis filter to generate an audio signal. Extracting the low frequency component of the audio signal and the high frequency information for obtaining the high frequency component of the audio signal,
The low frequency component is subjected to QMF analysis filtering to generate a low frequency sub-band signal,
Flatten the low band subband signal based on the energy and average energy of the low band subband signal;
Frequency-shifting the flattened low-frequency subband signal to a high frequency,
Based on the signal frequency shifted to the high frequency and the high frequency information, the gain of the signal frequency shifted to the high frequency is adjusted to generate a high frequency sub-band signal,
A signal processing method including a step of generating an audio signal by synthesizing the low frequency subband signal and the high frequency subband signal by a QMF synthesis filter. Extracting the low frequency component of the audio signal and the high frequency information for obtaining the high frequency component of the audio signal,
The low frequency component is subjected to QMF analysis filtering to generate a low frequency sub-band signal,
Flatten the low band subband signal based on the energy and average energy of the low band subband signal;
Frequency-shifting the flattened low-frequency subband signal to a high frequency,
Based on the signal frequency shifted to the high frequency and the high frequency information, the gain of the signal frequency shifted to the high frequency is adjusted to generate a high frequency sub-band signal,
A program that causes a computer to execute a process including a step of generating an audio signal by combining the low-frequency subband signal and the high-frequency subband signal with a QMF synthesis filter.

Images