KR101281945B1 - Apparatus and method for coding audio - Google Patents

Abstract

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an audio coder, and in particular, to remove a bass region of an input audio signal during encoding, to transmit bass signal information, and to generate artificial bass based on bass signal information transferred as additional information when decoding.

An audio encoding apparatus according to an embodiment of the present invention, in an encoder for encoding a digital audio signal, in the digital audio signal using an HPF (High Pass Filter) according to cutoff frequency information transmitted from an external control unit A bass remover for removing the bass signal, a fundamental frequency analyzer for detecting a fundamental frequency in the frequency domain using a fast fourier transform (FFT) result, and a bass information extractor for extracting bass signal related information using the fundamental frequency And a bit stream forming unit configured to output an encoded audio signal by configuring a bit stream using the bass signal related information and the bass-removed digital audio signal.

Bass, coding, encoder, decoder, audio coder

Description

Audio coding apparatus and method {APPARATUS AND METHOD FOR CODING AUDIO}

1A and 1B are block diagrams of a general perceptual encoder / decoder.

2a and 2b is a block diagram of a typical mp3 encoder and decoder,

3A and 3B are block diagrams of a perceptual encoder / decoder according to an embodiment of the present invention;

4A and 4B are block diagrams of an mp3 encoder / decoder according to an embodiment of the present invention;

5 and 6 are flowcharts illustrating an audio coding method according to an embodiment of the present invention.

The present invention relates to audio coders, and more particularly, to an apparatus and method for encoding / decoding digital audio signals.

In general, a perceptual audio coder is a system that compresses and reproduces sound by applying a perceptual model, which is an mp3 (MPEG Audio Layer-3) encoder / decoder or AAC (Advanced). Audio Coding) encoder / decoder, for example.

When the audio is compressed and reproduced by the perceptual audio coder, the focus is on how to effectively exclude unnecessary information and how much difference there is from the original sound when listening. Here, the general perceptual audio coder will be described, and then the detailed operation will be described using the MP3 encoder / decoder as an example.

1A and 1B show a block diagram of a general perceptual encoder / decoder. This is a general block diagram of an audio compression coder using a perceptual model.

First, referring to FIG. 1A, an encoder includes an analysis filter bank 105, a perceptual model application unit 120, a quantization and coding unit 110, and an encoding unit 115.

The audio signal is input to the analysis filter bank 105 and the perceptual model application unit 120, respectively.

When an audio signal is input, the analysis filter bank 105 applies a filter bank to an input digital audio sample, reconstructs it into a plurality of subband signals, and analyzes the frequency band. That is, when the audio signal is input, the analysis filter bank 105 divides a predetermined section in the frequency domain and separately applies the filter bank by applying the filter bank.

The perceptual model application unit 120 calculates an actual masking threshold to which the perceptual model is applied using the filter bank analysis result, and outputs the masking threshold to the quantization and coding unit 110. That is, the perceptual model application unit 120 analyzes the input audio signal so that an unnecessary part due to cognitive characteristics can be removed during audio compression.

The quantization & coding unit 110 quantizes and codes the audio signal to which the perceptual model is applied, into a digital signal with a predetermined error. For example, in the quantization process, the parts that are hard to hear are sparsely processed, and the parts that are hard to hear are finely processed. In other words, the quantization and coding unit 110 performs quantization and codes the spectral component so as to be below a masking threshold.

The encoder 115 forms the output of the quantization and coding unit 110 into a bit stream. The bit stream typically consists of quantized, coded spectral components, and side-information.

The decoder includes a decoder 135, an inverse quantizer 140, and a synthesis filter bank 145.

The decoder 135 decodes the input bit stream and outputs the decoded bit to the dequantizer 140.

The de-quantization unit 140 dequantizes the decoded bit stream output from the decoding unit 135 and outputs the dequantized bit stream to the synthesis filter bank 145.

The synthesis filterbank 145 applies a bank of synthesis filters to the output of the dequantizer 140 to generate a perceptually indistinguishable audio signal from the original audio signal. Output

2A and 2B are block diagrams of a general mp3 encoder and decoder.

The mp3 coder is a representative example of a perceptual encoding / decoding system. 32-band subband coding and Modified Discrete Cosine Transform (MDCT) are used for mp3 compression, and high efficiency compression is achieved using the psychoacoustic characteristics.

Referring to FIG. 2A, when a digital signal is input to an mp3 audio encoder, it is stored in an input buffer not shown. Digital data stored in the buffer is sequentially output to the filter bank 210.

The filter bank 210 reconstructs 32 subband signals by applying a filter bank to an input digital audio sample and analyzes each frequency band.

The MDCT 215 selects a window to be applied to the MDCT using perceptual energy.

In this case, the SMR (Signal to Masking Ratio) is obtained for each critical band, and if the perceptual energy is greater than or equal to the threshold, a short window is selected. If the perceptual energy is less than the threshold, the long window is determined. Choose.

After performing each MDCT, it is output to the quantization unit 230.

Fast Fourier transform (FFT) 220 performs an FFT on the 1024 (long window) and 256 (short window) of the input buffer.

The psychoacoustic model application unit 225 applies the psychoacoustic model to the result of the FFT 220 and outputs the psychoacoustic model to the quantization unit 230.

The quantization unit 230 quantizes the audio signal to which the psychoacoustic model is applied to a digital signal with a predetermined error. After quantization is performed, the Huffman coding unit 235 is output to the Huffman coding unit 235.

The Huffman encoding unit 235 performs Huffman coding on the output value of the quantization unit 230 and then outputs the bit stream forming unit 250. Here, the Huffman coding method is an algorithm widely used in audio, video, and various other fields. It considers the statistical frequency of data and assigns codes of different lengths according to the occurrence probability of the data, thereby compressing without loss of information. It is a method to carry out.

The bitstream formatting unit 240 configures a bitstream, and then outputs an audio signal in the form of a file after checking a cyclic redundancy checking (CRC).

The side-information coding unit 240 codes the additional information in the output value of the quantization unit 230 and then outputs the additional information to the bit stream forming unit 250.

The external control unit 245 provides information about the output device to the quantization unit 230.

Meanwhile, the mp3 decoder will be described with reference to FIG. 2B. The mp3 decoder may proceed in the reverse order of encoding of the mp3 encoder.

The coded audio signal in the form of a file is input to the frame unpacking unit 250.

The frame unpack unit 250 returns the compressed audio signal to its original state and transmits the compressed audio signal to the Huffman decoding unit 255 and the additional information coding unit 265, respectively.

The Huffman decoding unit 255 performs Huffman decoding on the unpacked spectral component that is the output value of the frame unpacking unit 250, and then outputs the Huffman decoding unit to the dequantization unit 260.

The additional information coding unit 265 decodes additional information from the unpacked data and outputs the decoded information to the dequantization unit 260.

The de-quantization unit 260 performs inverse quantization and outputs the inverse modified discrete cosine transform (IMDCT) 270.

The IMDCT 270 performs IMDCT on the output value of the inverse quantization unit 260 and then outputs it to iPQF 275.

The inverse poly-phase quadrature filter (iPQF) 275 synthesizes a filter bank for each subband and outputs a PCM type digital signal. It has the same function as the synthesis filter bank 145 described in FIG. 1B.

The conventional perceptual audio coder is an important problem of how to compress and reconstruct the same as the compressed source sound without considering the characteristics of the output device (for example, the speaker, built-in speaker, etc.) during compression and playback. However, it is necessary to apply a compression and reproduction method considering the characteristics of the output device to be used when the multimedia equipment is diversified and the output reproduction device according to the situation.

 In particular, the use of small speakers such as mobile phones, PDAs (Personal Digital Assistants), or PMPs (Portable Multimedia Players) can significantly reduce bass reproduction. Conventionally, it has to go through a post-processing process that uses an equalizer to forcibly emphasize low frequency bands or apply a missing fundamental technique to create a virtual bass.

  Among them, the missing fundamental technique is a psychoacoustic method that reflects human bass cognitive characteristics, allowing deeper and more natural bass listening compared to a simple equalizer. In other words, the missing fundamental technique is a technique related to human bass cognition characteristics. It is a method of making the harmonics of the low frequency region sound that cannot be physically produced by the playback device to make it feel like real bass. However, as much as human auditory illusion, detailed signal processing adjustment and a large amount of computation are required. Therefore, there is a problem in that it is difficult to apply to a mobile-multimedia device in which real-time processing is a priority with a small amount of calculation.

Accordingly, the present invention provides an audio coding apparatus and method for reducing the amount of computation by applying a missing fundamental technique to a recognition audio coder.

The present invention provides an audio coding apparatus and method for obtaining a better compression gain by applying a missing fundamental technique to a perceptual audio coder.

An audio encoding apparatus according to an embodiment of the present invention is an encoder for encoding a digital audio signal, the bass in the digital audio signal using a high pass filter (HPF) according to the cutoff frequency information transmitted from an external control unit A bass removal unit for removing a signal, a fundamental frequency analysis unit for detecting a fundamental frequency in a frequency domain using a fast fourier transform (FFT) result, a bass information extraction unit for extracting bass signal related information using the fundamental frequency, and And a bit stream forming unit configured to output a encoded audio signal by configuring a bit stream using the bass signal related information and the bass-removed digital audio signal.

An audio decoding apparatus according to an embodiment of the present invention is a decoder for decoding a digital signal, the harmonic generator for generating a harmonic of the fundamental frequency using the decoded bass signal-related information, and the BPF according to the cutoff frequency of the output device A BPF and a gain control unit for extracting the harmonics by applying a band pass filter and controlling gain values for each frequency to synthesize virtual bass, and a virtual bass adding unit for adding the virtual bass to the decoded bass signal related information. Include.

In an audio encoding method according to an embodiment of the present invention, in a method of encoding a digital audio signal, bass in the digital audio signal using a high pass filter (HPF) according to cutoff frequency information transmitted from an external control unit Removing a signal, detecting a fundamental frequency in a frequency domain using a fast fourier transform (FFT) result, extracting information about a bass signal using the fundamental frequency, And outputting an encoded audio signal by constructing a bit stream using the bass-removed digital audio signal.

In an audio decoding method according to an embodiment of the present invention, a method of decoding a digital signal, generating a harmonic of the fundamental frequency using the decoded bass signal-related information, and BPF (Band) according to the cutoff frequency of the output device And extracting the harmonics and controlling gains for each frequency to synthesize virtual bass, and adding the virtual bass to the decoded low signal information.

In the following description of the present invention, a detailed description of known functions and configurations incorporated herein will be omitted when it may make the subject matter of the present invention rather unclear. The following terms are defined in consideration of the functions of the present invention, and may be changed according to the intentions or customs of the user, the operator, and the like. Therefore, the definition should be based on the contents throughout this specification.

The present invention performs a fundamental frequency analysis (fundamental frequency analysis) using the FFT result of the psychoacoustic block in the encoder stage in order to apply the missing fundamental technique to the conventional perceptual audio coder. In addition, according to the characteristics of the output device to which the encoding sound source is to be reproduced, the low sound region of the input audio signal is removed in advance, and the low band subband is omitted.

The decoder has a feature of generating artificial bass based on bass signal information (fundamental frequency, magnitude, low band average energy, etc.) transferred as additional information.

3A and 3B are block diagrams of a perceptual encoder / decoder according to an embodiment of the present invention.

Referring first to the encoder shown in FIG. 3A, the following blocks have been added to a conventional perceptual audio coder (see FIG. 1A). The added blocks will be described.

The bass subtraction unit 310 removes the bass signal using the HPF (High Pass Filter) according to the cutoff frequency information transmitted from the external controller. At this time, the cutoff frequency is based on the cutoff frequency of the output device. When using an earphone or a high-performance speaker, a low cutoff frequency is set, and a high cutoff frequency is set in the case of a mobile phone micro speaker having poor reproduction characteristics. As such, by removing the bass signal, the dynamic range of the input signal may be lowered, and the subband of the low band may be omitted when the MDCT is performed in the MDCT. This, in turn, has the advantage of lowering the bit rate during compression compared to the prior art.

The bass signal analysis unit 320 detects the bass signal using the result of applying the perceptual model at the encoder stage to apply the missing fundamental technique to the conventional perceptual audio coder, and then outputs it to the encoding unit 115. do.

Looking at the decoder shown in Figure 3b, it is as follows.

The virtual bass generation unit 330 generates virtual bass using information on the decoded bass signal.

The virtual bass adding unit 340 outputs a bit stream by adding the virtual bass generated by the virtual bass generator 330 to the decoded audio signal.

4A and 4B illustrate block diagrams of an mp3 encoder / decoder according to an embodiment of the present invention.

4A, the following blocks have been added to the conventional mp3 encoder. The added blocks will be described.

The bass removal unit 410 removes the bass signal using the HPF (High Pass Filter) according to the cutoff frequency information transmitted from the external controller. At this time, the cutoff frequency is based on the cutoff frequency of the output device. When using an earphone or a high-performance speaker, a low cutoff frequency is set, and a high cutoff frequency is set in the case of a mobile phone micro speaker having poor reproduction characteristics. As such, by removing the bass signal, the dynamic range of the input signal can be lowered, and the subband of the low band can be omitted when the MDCT is performed in the MDCT. This in turn has the advantage of lowering the bit rate during compression than the prior art.

The basic frequency analyzer 420 performs an FFT on the input audio signal in order to apply the psychoacoustic model in the perceptual audio coder. The basic frequency analyzer 420 is used for analysis and has no additional role for compression. The fundamental frequency analyzer 420 of the present invention detects the fundamental frequency required by the missing fundamental technique using the FFT result and analyzes the harmonics. When applied to a post-decoding process as in the prior art, if an additional calculation is required before the final output and the FFT is performed to find the fundamental frequency, the operation is not only repeated but also takes a large amount of computation. However, if the FFT results are pre-analyzed at the encoder stage, the computation can be significantly reduced and more accurate analysis can be performed in the frequency domain.

Bass Info. Extraction unit 430 calculates the location and energy of the fundamental frequency and the average energy of the low frequency band below the cutoff frequency using the fundamental frequency detected by the fundamental frequency analysis 420. This is encoded by the additional information coding unit 240 and added to the bit stream forming unit 250. Based on this information, the decoder produces artificial bass signals. In this case, adding harmonics using the missing fundamental technique can make the output device sound like it is not physically playable. This is called virtual bass.

In the decoder illustrated in FIG. 4B, the following block is added to the block diagram of FIG. 2B. The added blocks will be described.

The virtual bass synthesis unit 440 includes a harmonic generator 442 and a BPF & gain controller 444.

The harmonics generation unit 442 uses the information about the decoded bass signal (base frequency position, magnitude in the fundamental frequency, average energy of the bass band) for the fundamental frequency. Create a harmonic

The BPF & gain control unit 444 extracts the harmonics by applying the BPF according to the cutoff frequency of the output device and controls the gain value for each frequency so as not to cause distortion, thereby adding a virtual bass. Output at 450.

The virtual bass adding unit 450 mixes the virtual bass with the decoded audio signal according to the selection of the controller 460.

The control unit 460 controls the BPF and the gain value for the harmonic extraction according to the cutoff frequency of the output device (earphone, Hi-Fi speaker, micro speaker, etc.), and performs the mixing before the final output and gain control during the mixing. do.

5 and 6 are flowcharts illustrating an audio coding method according to an exemplary embodiment of the present invention. 5 is a flowchart illustrating an encoding operation in an encoder according to an embodiment of the present invention.

The encoder receives the PCM data in step 501. In step 503, the encoder receives the cutoff frequency and the bass removal information from the external controller, and determines whether to remove the bass from the received PCM data. At the same time, the encoder proceeds to step 513. A description of the step 513 will be described below.

When removing the bass, the encoder proceeds to step 505 to remove the bass. At this time, the encoder removes the bass using the HPF according to the cutoff frequency information transmitted from the external controller. After removing the bass, go to step 507.

However, if the bass is not removed, the encoder applies a filter bank to the PCM data input in step 507 to filter and reconstruct the 32 subband signals, and analyzes each frequency band. The encoder then selects the window applied to the MDCT in step 509. In this case, the SMR (Signal to Masking Ratio) is obtained for each critical band, and if the perceptual energy is greater than or equal to the threshold, a short window is selected. If the perceptual energy is less than the threshold, the long window is determined. Choose.

In step 511, the encoder quantizes the audio signal to which the psychoacoustic model is applied, into a digital signal with a predetermined error. After quantization is performed, Huffman coding is performed. The encoder configures the bit stream in step 521, and then outputs an audio signal in the form of a file after performing a cyclic redundancy checking (CRC) check.

Meanwhile, in step 513, which occurs simultaneously with step 503, the encoder performs a fast fourier transform (FFT) on 1024 (long window) and 256 (short window) for the input buffer, and then examines the FFT result in step 515. Apply the acoustic model.

After that, the encoder detects the fundamental frequency using the FFT result applied to the psychoacoustic model in step 517 to analyze the harmonic characteristics. In other words, the encoder uses the FFT results to find the fundamental frequencies required by the missing fundamental technique and analyze the harmonics. In step 519, the encoder calculates the position and magnitude of the fundamental frequency and extracts the bass signal information by calculating the average energy for the bass region below the cutoff frequency. The extracted basal signal information is coded through the additional information coding unit 240 and added to the bit stream forming unit 250. The bit stream forming unit 250 configures a bit stream, and then outputs an audio signal in the form of a file after a CRC check for error checking.

6 is a flowchart illustrating a decoding operation in a decoder according to an embodiment of the present invention.

First, in step 601, the decoder unpacks the coded audio data in the form of a file. The decoder then decodes the unpacked data Huffman for the spectral component in step 603. At the same time, the decoder decodes the bass signal information in step 613. Steps 613 and below will be described below.

After step 603, the decoder performs dequantization in step 605 and then performs IMDCT in step 607. The decoder then synthesizes a filter bank for each subband in step 609 and then determines whether to add virtual bass in step 611. If no virtual bass is added, the decoder terminates the decoding operation. However, when adding virtual bass, the decoder proceeds to step 619 to add the virtual bass and ends the decoding operation.

On the other hand, in step 613, which proceeds simultaneously with step 603, the decoder decodes the additional information of the bass signal from the unpacked additional information.

In operation 615, the decoder generates a harmonic of the fundamental frequency by using the information about the decoded bass signal (base frequency position, magnitude in the fundamental frequency, and average energy of the bass band). In step 617, the decoder extracts the harmonics by applying the BPF according to the cutoff frequency of the output device and controls the gain value for each frequency such that distortion does not occur.

In operation 619, the decoder adds the virtual bass to the virtual bass and the decoded audio signal according to the selection of the controller 460.

 Meanwhile, in the detailed description of the present invention, specific embodiments have been described, but it will be apparent to those skilled in the art that various modifications are possible without departing from the scope of the present invention.

In the present invention that operates as described in detail above, the effects obtained by the representative ones of the disclosed inventions will be briefly described as follows.

The present invention can reduce the amount of calculation by about 1/3 compared to the prior art during decoding.

In addition, since the present invention omits a band that cannot be physically reproduced by an output device during encoding, the compression rate can be increased in terms of bit rate. At least 1/32 of the original data can be reduced and the dynamic range of the input signal can be reduced, so additional data compression can be expected.

In addition, the present invention can improve the performance of the bass compensation function by detecting the fundamental frequency more accurately than the conventional missing fundamental technique applied to the post-processing by using the FFT information used when applying the psychoacoustic model in the encoder stage.

The present invention can reduce the post-processing calculation amount for low tone compensation by passing a part of the missing fundamental technique, which takes up a large amount of computation, to the encoder. As a result, the missing fundamental technique can be applied to an embaded system (mobile-multimedia device), which was difficult to apply in the related art.

In addition, according to the present invention, by applying a missing fundamental, a low bass signal that cannot be physically reproduced in a poor output device can be heard as a virtual bass. Better bass reproduction is possible than conventional methods using equalizers.

Claims (8)

An encoder for encoding a digital audio signal, A bass removing unit for removing bass signals from the digital audio signal using a high pass filter (HPF) according to cutoff frequency information transmitted from an external controller; A fundamental frequency analyzer for detecting a fundamental frequency in a frequency domain by using a fast fourier transform (FFT) result; A bass information extraction unit for extracting bass signal related information using the fundamental frequency; And a bit stream forming unit configured to output a encoded audio signal by constructing a bit stream using the bass signal related information and the bass-removed digital audio signal. The method of claim 1, The bass signal related information, And at least one of a position, an energy of the fundamental frequency, and a low frequency average energy below a cutoff frequency. A decoder for decoding a digital signal, A harmonic generator for generating harmonics with respect to the fundamental frequency using the decoded bass signal related information; A BPF and gain control unit which extracts the harmonics by applying a band pass filter (BPF) according to a cutoff frequency of an output device and synthesizes virtual bass by controlling gain values for each frequency; And a virtual bass adder configured to add the virtual bass to the decoded bass signal related information. The method of claim 3, The bass signal related information, And at least one of a position, an energy of the fundamental frequency, and a low frequency average energy below a cutoff frequency. In the method of encoding digital audio signals, Removing a bass signal from the digital audio signal using a high pass filter (HPF) according to cutoff frequency information transmitted from an external controller; Detecting the fundamental frequency in the frequency domain using the fast fourier transform (FFT) result, Extracting information related to a bass signal using the fundamental frequency; And constructing a bit stream using the bass signal related information and the bass-removed digital audio signal to output an encoded audio signal. The method of claim 5, The bass signal related information, And at least one of a position, an energy of the fundamental frequency, and a low frequency average energy below a cutoff frequency. In the method of decoding a digital signal, Generating a harmonic of a fundamental frequency using information about the decoded bass signal; Extracting the harmonics by applying a band pass filter (BPF) according to a cutoff frequency of an output device and synthesizing a virtual bass by controlling a gain value for each frequency; And adding the virtual bass to the decoded bass signal related information. The method of claim 7, wherein The bass signal related information, And at least one of a position, an energy of the fundamental frequency, and a low frequency average energy below a cutoff frequency.

Images