JP5392057B2 - Audio processing apparatus, audio processing method, and audio processing program - Google Patents

Description

  The present invention relates to a voice processing apparatus, a voice processing method, and a voice processing program for analyzing a digital voice signal and processing the digital voice signal using the analysis result.

  In recent years, due to advances in audio coding technology, it has become possible to reduce the file size while maintaining the sound quality of music recorded on CDs (Compact Discs) as much as possible. For example, a large amount of music can be recorded and carried.

  However, the above-described speech coding technology cuts out high-frequency speech signals that are not normally heard using human auditory characteristics, or thins out sound data that cannot be heard due to the masking effect. In comparison, the sound will be less stretched, spread, dynamic range and glossy. Therefore, a technique for improving the sound quality of a digital audio signal compressed by an audio encoding technique has been developed.

  For example, the maximum and minimum values of a digital audio signal are identified, the number of samples from the minimum value to the maximum value, or from the maximum value to the minimum value is counted, and for each sample excluding the maximum value and the minimum value, A technique is disclosed in which a difference from a sample value is calculated, a coefficient corresponding to the number of samples is multiplied, and the multiplication result is added to or subtracted from a sample position close to a maximum value or a minimum value (for example, Patent Document 1).

  Similarly, the number of samples between extreme values is counted, the difference between the maximum value or the minimum value and the value one sample before is calculated, and this is multiplied by a coefficient corresponding to the number of samples. There is also known a technique of adding or subtracting directly to or from a sample position close to a maximum value or minimum value (for example, Patent Document 2).

Japanese Patent No. 3401171 Japanese Patent No. 3659489

  In the techniques of Patent Documents 1 and 2 described above, the interval from the minimum value to the maximum value and the interval from the maximum value to the minimum value are controlled independently, and the coefficient to be multiplied according to the number of samples in each interval, The sample position to be added / subtracted is determined. Therefore, appropriate sound quality improvement processing can be performed for each section corresponding to a half cycle even for irregular changes in the digital audio signal.

  On the other hand, since there is no guarantee that the frequency of each signal included in the digital audio signal is equal to the sampling frequency, even if the digital audio signal is a regular sine wave, if the frequency is different from the sampling frequency, it is half There may occur a case where the number of samples per period, that is, the number of samples from the minimum value to the maximum value is different from the number of samples from the maximum value to the minimum value.

  Thus, if the number of samples between extreme values is different, the coefficient to be multiplied and the sample position to be added or subtracted will change, and the amount of improvement in sound quality will differ every half cycle, so even for regular sine waves In some cases, the sound quality improvement processing is biased, and the sound quality improvement effect is not sufficiently exhibited.

  In view of such a problem, the present invention provides a sound processing apparatus capable of achieving uniform sound quality improvement by uniformly performing sound improvement processing on each signal included in a digital sound signal, An object is to provide a voice processing method and a voice processing program.

In order to solve the above-described problem, the audio processing apparatus of the present invention performs frequency analysis of an input digital audio signal, and removes one or more fundamental wave signals and one or more fundamental wave signals from the digital audio signal. A signal separation unit that separates the residual signal into a residual signal, and a correction signal addition unit that generates a correction signal whose amplitude is expanded for each of one or a plurality of fundamental wave signals and adds the correction signal to the fundamental signal. A residual signal adding unit that adds a residual signal to one or a plurality of fundamental signals to which correction signals are added, respectively , wherein the one or the plurality of fundamental signals are a plurality of fundamental signals having different frequencies. The signal separation unit obtains a difference signal when a plurality of fundamental wave signals having the same frequency as that of one or more fundamental wave signals are subtracted from the digital audio signal, respectively, and the difference signal is decremented in ascending order of energy of the difference signal. Characterized that you separate the digital audio signal into a one or more of the fundamental wave signal and the residual signal by sequentially subtracting one or more of the fundamental wave signal from the barrel audio signal.

The audio processing device cuts out a digital audio signal in a predetermined frame unit, generates a digital audio signal for each predetermined frame, and inputs the input digital audio signal in a frame unit into a digital audio of an adjacent frame. An overlap synthesis unit that synthesizes the signals so that part of the signals overlap, and the digital audio signal input to the signal separation unit is a digital audio signal divided into predetermined frames generated by the framing unit The digital audio signal in units of frames input to the overlap synthesis unit may be input from the residual signal addition unit.

  The one or more fundamental wave signals described above may be signals represented by a predetermined frequency and the amplitudes of a sine wave and a cosine wave having the predetermined frequency.

  The correction signal adding unit may generate a correction signal according to the frequency of each of the one or more fundamental wave signals, the amplitude of the sine wave, and the amplitude of the cosine wave. Specifically, the correction signal adding unit refers to a correction table in which the frequency of the fundamental wave signal and the value of the correction signal at each sample position of the sine wave and cosine wave having an amplitude of 1 are associated in advance. Alternatively, the value of the correction signal at each sample position of the sine wave and cosine wave whose amplitude is 1 is extracted according to the frequency of each of the plurality of fundamental wave signals, and the amplitude and cosine of the sine wave of each of the one or more fundamental wave signals. The correction signal may be generated by multiplying the amplitude of the wave.

In order to solve the above-described problem, the audio processing method of the present invention performs frequency analysis of an input digital audio signal, and each of a plurality of fundamental wave signals having the same frequency as one or a plurality of fundamental wave signals having different frequencies. A difference signal when subtracted from the digital audio signal alone is obtained, and one or more fundamental wave signals are sequentially subtracted from the digital audio signal in order of increasing energy of the difference signal, so that the digital audio signal is converted into one or more fundamental wave signals. Separated into residual signals excluding one or a plurality of fundamental signals, a correction signal is generated for each of the one or more fundamental signals so that the absolute value of the amplitude is expanded and added to the fundamental signal The residual signal is added to one or a plurality of fundamental wave signals to which the correction signals are added.

In order to solve the above problems, a sound processing program of the present invention performs frequency analysis of a digital sound signal input to a computer, and a plurality of fundamental waves having the same frequency as one or a plurality of fundamental wave signals having different frequencies. A difference signal is obtained when each signal is subtracted from the digital audio signal alone, and one or more fundamental wave signals are sequentially subtracted from the digital audio signal in order of increasing energy of the difference signal to thereby obtain one or more basic digital audio signals. A signal separation step for separating a wave signal and a residual signal from which one or more fundamental wave signals are removed, and a correction signal that increases the absolute value of the amplitude for each of the one or more fundamental wave signals And a correction signal generating step for adding to the fundamental signal, and a residual signal for adding the residual signal to one or a plurality of fundamental signals to which the correction signal has been added. Characterized in that to execute a signal adding step.

  As described above, according to the present invention, it is possible to achieve uniform sound quality improvement by uniformly performing sound improvement processing on each signal included in a digital sound signal.

It is explanatory drawing for demonstrating the utilization condition of a speech processing unit. It is a functional block diagram for demonstrating the schematic structure of a speech processing unit. It is explanatory drawing for demonstrating the production | generation process of the frame signal in a framing part. It is explanatory drawing which showed an example of the predetermined number of frequency used as a frequency analysis candidate. It is a functional block diagram for demonstrating the more specific structure of a correction signal addition part. It is a coefficient table showing the relationship between the number of samples and coefficients. It is explanatory drawing for demonstrating operation | movement of the sound quality improvement process by the correction signal addition part. It is explanatory drawing for demonstrating operation | movement of the sound quality improvement process by the correction signal addition part. It is explanatory drawing for demonstrating the sound quality improvement process in a correction signal addition part. It is explanatory drawing for demonstrating operation | movement of an overlap synthetic | combination part. It is a functional block diagram showing a typical example of a computer. It is the flowchart which showed the whole flow of the speech analysis synthesis method.

  Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. The dimensions, materials, and other specific numerical values shown in the embodiments are merely examples for facilitating the understanding of the invention, and do not limit the present invention unless otherwise specified. In the present specification and drawings, elements having substantially the same function and configuration are denoted by the same reference numerals, and redundant description is omitted, and elements not directly related to the present invention are not illustrated. To do.

(Speech processor 100)
FIG. 1 is an explanatory diagram for explaining a usage state of the speech processing apparatus 100. The audio processing apparatus 100 acquires a digital audio signal through broadcast waves from the broadcast station 102, through the communication network 106 from the content server 104, or directly from the storage medium 108, and adds a high frequency component to the digital audio signal. To improve the sound quality of digital audio signals. The user can listen to the improved digital audio signal directly from the audio processing device 100 or by transferring it to a playback device 110 such as a portable audio player or a mobile phone.

  Further, the content server 104 may include the audio processing device 100. In this case, the audio signal to which the high frequency component is added by the audio processing device 100 of the content server 104 is used as a personal computer, a portable audio player, or a mobile phone. To the playback device 110 through the communication network 106.

  Further, the playback device 110 such as a portable audio player or a mobile phone may include the sound processing device 100. In this case, a digital audio signal distributed from the content server 104 through the communication network 106 is reproduced with a high frequency component added by the audio processing device 100 of the reproduction device 110 such as a portable audio player or a mobile phone.

  Digital audio signals that can be acquired by the audio processing apparatus 100 include audio signals based on CD and DVD (Digital Versatile Disk) standards, MPEG (Moving Picture Expert Group) -2, AAC (Advanced Audio Coding), and HE-. Audio signal whose frequency band is narrowed by audio encoding processing such as AAC (High Efficiency-AAC), ATRAC (Adaptive TRansform Acoustic Coding), MP3 (MPEG Audio Layer-3), WMA (Windows (registered trademark) Media Audio) Including. Here, as an example of the input digital audio signal, each function of the audio processing apparatus 100 will be described by taking a digital audio signal (CD standard) having a sampling frequency fs = 44.1 kHz and a quantization bit number of 16 bits.

  FIG. 2 is a functional block diagram for explaining a schematic configuration of the sound processing apparatus 100. The speech processing apparatus 100 includes a framing unit 120, a signal separation unit 122, a correction signal addition unit 124, a residual signal addition unit 126, and an overlap synthesis unit 128.

  The framing unit 120 sequentially cuts out the digital audio signal acquired by the audio processing apparatus 100 in a predetermined frame unit (predetermined number of samples) as a processing unit, and a digital audio signal in frame units (hereinafter simply referred to as a frame signal). Is generated.

  FIG. 3 is an explanatory diagram for explaining a frame signal generation process in the framing unit 120. As shown in FIG. 3, when one continuous digital audio signal is input, the framing unit 120 firstly selects a part of the digital audio signal A that is delimited by a predetermined length from the input digital audio signal. Only the frame signal 0 is cut out to generate the frame signal 0. At this time, since there is no digital audio signal before the digital audio signal A, the frame signal 0 having a predetermined number of samples including the digital audio signal A is formed from a null value and the digital audio signal A as shown in FIG. Is done. The framing unit 120 temporarily stores a rear signal A ′, which is data of a predetermined length from a predetermined position to the tail of the digital audio signal A, in a buffer (not shown) for the next frame signal. .

  Subsequently, the framing unit 120 cuts out the digital audio signal B in accordance with the digital audio signal that is continuously input, and connects the rear signal A ′ and the digital audio signal B that have been held in that order. Then, a frame signal 1 having a predetermined sample length is generated. Thereafter, the framing unit 120 repeats the generation of the frame signal such that the frame signal 2 is generated from the rear signal B ′ of the digital audio signal B and the digital audio signal C cut out next.

  Therefore, the frame signal generated by the framing unit 120 partially overlaps with the preceding and following frame signals. For example, the data corresponding to the rear signal A ′ overlaps between the frame signal 0 and the frame signal 1. In the subsequent signal separation unit 122, correction signal addition unit 124, residual signal addition unit 126, and overlap synthesis unit 128, processing is performed on the frame signal (digital audio signal in frame units). In addition, here, the length of the overlapped rear signals A ′, B ′,... Is illustrated as 1/3 of the length of a frame signal having a predetermined number of samples. However, it can be an arbitrary numerical value of 1/2 or less.

  In the present embodiment, the frame signals generated in this way are combined again while being overlapped by an overlap combining unit 128 described later. Such an overlap portion makes it possible to ensure the continuity of the digital audio signal, and also ensure the continuity of the correction signal that forms a high frequency component newly generated based on the embodiment. In this way, it is possible to avoid the influence of edges (edges) caused by cutting out the frame signal, and to obtain a stable sound quality improvement effect.

  Then, the framing unit 120 sequentially transmits the generated frame signal to the signal separation unit 122.

  The signal separation unit 122 performs frequency analysis of the frame signal divided into predetermined frames received from the framing unit 120, and removes the frame signal from one or more fundamental wave signals and one or more fundamental wave signals. Separated into residual signals. In the present embodiment, the signal separation unit 122 separates the fundamental signal and the residual signal using Generalized Harmonic Analysis (GHA).

  Such general harmonic analysis has a higher computational load compared to Fast Fourier Transform (FFT), which is widely used as a frequency analysis method, but (1) has higher frequency analysis accuracy than Fast Fourier Transform. (2) It is advantageous in that noise can be suppressed and (3) a waveform other than the frame signal to be analyzed can be predicted.

  When the frequency analysis of the frame signal is performed using the fast Fourier transform, the frame signal is handled as a periodic function in units of frames, so that a discontinuous frequency component is generated at the end, and the digital audio signal that becomes the original signal A new frequency component not included in is detected. Furthermore, if a window function is applied to ensure the continuity of the end of the frame signal, the frequency analysis result of the fast Fourier transform is always affected by the window function.

  On the other hand, in the general harmonic analysis of this embodiment, since an appropriate combination of sine wave and cosine wave that minimizes residual energy is derived from the frame signal, frequency analysis is performed with high frequency resolution that does not depend on time resolution. Can be carried out. As described above, it is most desirable that the signal separation unit 122 separates the fundamental wave signal and the residual signal using general harmonic analysis, but the present invention is not limited to this, and various frequency analysis methods may be used. it can.

In accordance with such general harmonic analysis, the signal separation unit 122 first determines a predetermined number of different frequencies f _k (k is an integer) as frequency analysis candidates based on the sampling frequency fs. Then, the fundamental wave signal b _{k [i]} (i is 0 to L-1 integer, L is the number of samples of the frame signal) of the determined predetermined number of frequency f _k and singly by subtracting from the frame signal The difference signal e _k [i] is obtained, and the energy E _k of the difference signal is derived from the sum of squares thereof.

The signal separation unit 122 acquires information on the sampling frequency of the digital audio signal extracted when a decoder (not shown) decodes the digital audio signal, and a predetermined number of phases that are frequency analysis candidates according to the sampling frequency. Different frequencies f _k (k is an integer) may be determined. However, when the audio processing apparatus 100 according to the present embodiment is used in a playback apparatus in which the sampling frequency of the input digital audio signal is always constant, such as a CD player, the signal separation unit 122 does not necessarily have the sampling frequency information. There is no need to acquire.

FIG. 4 is an explanatory diagram showing an example of a predetermined number of frequencies that are frequency analysis candidates. Here, as the predetermined number of frequencies, a frequency having the same number of samples in a half cycle before and after the extreme value in the waveform of the frequency is selected. In this embodiment, since the sampling frequency fs is 44.1 kHz, the frequency f _k in which the number of samples is the same in the half cycle before and after the extreme value is equal to half the sampling frequency fs and the number of samples in the half cycle. A value fs / 2 / FS divided by FS (FS is an integer) is obtained.

However, since the frequency component of the frequency f ₁ (22.05 kHz) of FS = 1 is not included in the frame signal x ₀ [i] (i is an integer of 0 to L−1) to be processed according to the sampling theorem, As shown in FIG. 4, the frequency f _k is limited to the frequency at which the number of samples FS = 2, 3, 4,. In this embodiment, nine different frequencies f _{2 to} f ₁₀ with the number of samples FS = 2, 3, 4,. The reason why the frequency _{fk is set} to the frequency at which the number of samples is the same in the half cycle before and after the extreme value will be described later.

…（数式１） Further, the fundamental wave signal b _k [i] having the frequency f _k can be expressed by Equation 1. However, i is 0 to L-1, and k is 2, 3, 4,.

... (Formula 1)

…（数式２）

…（数式３） The signal separation unit 122 selects the frequencies shown in FIG. 4 in ascending order of FS, and the amplitude S (f _k ) of the sine wave of the fundamental wave signal b _k [i] with respect to the frame signal x ₀ [i] to be processed. Is derived using Equation 2, and the amplitude C (f _k ) of the cosine wave is derived using Equation 3. However, k is 2, 3, 4,...

... (Formula 2)

... (Formula 3)

…（数式４） The fundamental wave signal b _k [i] is obtained by substituting the amplitude S (f _k ) and the amplitude C (f _k ) derived in this way into Equation 1, and the frame signal x ₀ [i] to be processed is obtained. Thus, the fundamental signal b _k [i] is subtracted individually as shown in Equation 4 to obtain a differential signal e _k [i].

... (Formula 4)

…（数式５） Then, the energy E _k of the difference signal e _k [i] is derived by the sum of squares as shown in Equation 5, and temporarily stored in association with the frequency f _k .

... (Formula 5)

Here, as the energy E _{k of} the predetermined number of derived difference signals e _k [i] is smaller, the fundamental signal b _k [i] of the frequency f _{k is} the frame signal x ₀ [i] to be processed. It represents that the occupation rate (degree) included in is high. The signal separation unit 122 uses the energy E _k of the differential signal e _k [i] for all nine frequencies f _{k of} fs / 2 / FS (FS = 2, 3, 4,... 10) shown in FIG. calculate.

The reason why the energy E _k of the differential signal e _k [i] is individually determined is to minimize the final residual signal except for all of the one or more fundamental wave signals b _k [i]. This is because, under general harmonic analysis, it is necessary to preferentially separate the fundamental wave signal b _k [i] having a high occupation rate from the frame signal x ₀ [i]. Therefore, the signal separation unit 122 rearranges the nine frequencies f _k in ascending order of the differential energy E _k , that is, in descending order of the occupation ratio in the fundamental wave signal b _k [i].

…（数式６） Then, the signal separation unit 122, the nine corresponding to the frequency _{f k} nine fundamental signal _b k [i], the order of the frequency _{f k} rearranged, a original signal frame signal _x 0 [i ] In turn. However, in the above-described step of deriving the differential signal e _k [i], the fundamental signal b _k [i] is subtracted from the frame signal x ₀ [i], which is the original signal, every time. When one fundamental wave signal b _k [i] is subtracted from the frame signal x ₀ [i], the amplitude of the next fundamental wave signal b _k [i] with respect to the residual signal d [i] after the subtraction. S (f _k ) and amplitude C (f _k ) are derived again using Equations 2 and 3, and the fundamental wave signal b _k [i] is subtracted. Therefore, depending on the order of subtraction, the amplitude S (f _k ) and amplitude C (f _k ) of the fundamental wave signal b _k [i] change. Table sine wave and cosine wave having a predetermined frequency f _k is the fundamental wave signal b _{k [i]} are sequentially subtracted from the frame signal x _{0 [i]} the fundamental wave signal b _{k [i]} used in the rearrangement In common, only the amplitudes of the sine wave and cosine wave are different. Fundamental signal b _k used for the rearrangement _[i] will not be used and rearrangement is complete, the amplitude S (f _k) and the amplitude C (f _k) has changed the fundamental wave signal b _{k [i]} is The final fundamental wave signal b _k [i] is also used in the subsequent processing. The residual signal d [i] is derived through the subtraction of the fundamental wave signal b _k [i]. Therefore, the residual signal d [i] can be expressed as Equation 6. However, i is 0 to L-1, and k is 2, 3, 4,.

... (Formula 6)

In this way, the fundamental signal b _k [i] having a high occupation rate is sequentially separated from the frame signal x ₀ [i], and the energy of the residual signal d [i] is gradually reduced.

Thus, the frame signal x ₀ by the configuration which separates preferentially from the fundamental wave signal is high occupancy b _{k [i]} in the _[i], a frame signal x _{0 [i]} one or more of the fundamental wave signal b _k The combination of [i] can be appropriately expressed, and the residual signal d [i] can be minimized.

Here, since a predetermined number (here, 9) of frequencies _fk as frequency analysis candidates are uniquely obtained as shown in FIG. 4 with respect to the sampling frequency fs (for example, 44.1 kHz), the frequency _fk is determined according to the sampling frequency fs. A fundamental wave table in which a predetermined number of frequencies f _k and fundamental wave signals b _k [i] are uniquely associated can be created in advance. However, the fundamental wave table only shows the values of the sine wave and cosine wave in each sample i when the amplitude S (f _k ) and the amplitude C (f _k ) are set to predetermined values (for example, 1). The separating unit 122 multiplies the amplitude S (f _k ) and the amplitude C (f _k ) to derive the fundamental wave signal b _k [i]. Such a fundamental wave table can reduce the calculation load. Such a fundamental wave table may be held in a memory (not shown) or may be acquired from the communication network 106.

Signal separator 122, gradually subtracting the fundamental wave signal b _{k [i]} to continue the sorted order, by subtracting the fundamental wave signal b _{k [i]} for all frequencies f _k, which is prepared as a frequency analysis candidates When finished, the residual signal d [i] is transmitted to the residual signal adding unit 126 as a final residual signal.

Here, even if the fundamental signal b _k [i] for all the frequencies f _k prepared as frequency analysis candidates is not subtracted, if the energy of the residual signal d [i] is sufficiently small, For example, if the residual signal d [i] is equal to or lower than a predetermined energy, the separation of the fundamental signal b _k [i] is stopped at that time, assuming that the frame signal x ₀ [i] can be sufficiently separated. Residual signal d [i] is transmitted to residual signal adding section 126.

At this time, the fundamental wave signal b _k [i] is a signal represented by a predetermined frequency, an amplitude of a sine wave having a predetermined frequency, and an amplitude of a cosine wave. Not the wave signal b _k [i] itself, but parameters such as frequency information indicating the frequency of the fundamental wave signal b _k [i], amplitude information of the sine wave component, amplitude information of the cosine wave component, and the fundamental wave signal b _k. The number information of [i] is transmitted to the correction signal adding unit 124. With this configuration, the access load between the signal separation unit 122 and the correction signal addition unit 124 can be significantly reduced.

Further, as shown in Expression 6, the signal separation unit 122 adds a residual signal d [i] obtained by removing all the target fundamental wave signals b _k [i] from the frame signal x ₀ [i]. To the unit 126.

In the present embodiment, as will be described later, only the fundamental wave signal b _k [i] is subjected to the sound quality improvement process, and is not applied to the residual signal d [i]. However, since the residual signal d [i] is a signal that can be ignored as the amount of energy, the sound quality of the frame signal x ₀ [i] that is the original signal is improved without performing the sound quality improvement process on the residual signal d [i]. There is no effect on the level, but rather the processing load for applying the sound quality improvement processing to the residual signal d [i] can be effectively utilized for other processing.

Further, as shown in FIG. 4, the frequency f _k of the fundamental wave signal b _k [i] is set to a frequency f _k = fs / 2 / FS at which the number of samples is the same in a half cycle before and after the extreme value. Thus, the sine wave having the same number of samples in the half period before and after the extreme value of the frame signal x ₀ [i] (one or more fundamental wave signals b _k [i]) excluding the residual signal d [i]. In the same sine wave or cosine wave, there is no problem that the coefficient to be multiplied or the sample position to be added or subtracted is different. Furthermore, since the fundamental wave signal b _k [i] is formed only by a sine wave and a cosine wave having an initial phase 0, a correction signal can be uniformly added to the frame signal x ₀ [i]. Therefore, it is possible to achieve uniform sound quality improvement.

  The correction signal adding unit 124 generates a correction signal for each of one or a plurality of fundamental wave signals separated by the signal separation unit 122 so that the absolute value of the amplitude centered on the sound pressure 0 is expanded to generate a fundamental wave. Add to signal.

  FIG. 5 is a functional block diagram for explaining a more specific configuration of the correction signal adding unit 124. FIG. 6 is a coefficient table showing the relationship between the number of samples and the coefficients. These are explanatory drawings for explaining the operation of the sound quality improvement processing by the correction signal adding unit 124. The correction signal adding unit 124 includes an extreme value specifying unit 150, a sample number counting unit 152, a correction signal generating unit 154, a delay unit 156, and an adding unit 158. The coefficient table may be held in a memory (not shown) or may be acquired from the communication network 106. First, the basic operation of the sound quality improvement process executed by the correction signal adding unit 124 will be described.

The extreme value specifying unit 150 specifies the maximum value and the minimum value of the frame signal x ₀ [i] (one or a plurality of fundamental wave signals b _k [i]) received by the correction signal adding unit 124. Specifically, the extreme value specifying unit 150 sequentially compares the values in the respective samples of the frame signal x ₀ [i], and when the value changes from a state where the value is increasing or a state where there is no increase / decrease, the extreme value specifying unit 150 The value in the sample immediately before turning is set to the maximum value, and when the value is decreasing or the state in which there is no increase / decrease is changed to the increase, the value in the sample immediately before starting the increase is set to the minimum value.

  The sample number counting unit 152 counts the number of samples from an arbitrary extreme value (maximum value or minimum value) to the next extreme value, that is, the number of samples from the maximum value to the minimum value, or from the minimum value to the maximum value. .

The correction signal generation unit 154 generates a correction value such that the absolute value of the amplitude of the digital audio signal is expanded by multiplying the change amount between predetermined samples in the frame signal x ₀ [i] and a coefficient of 1, A correction signal is generated at a predetermined sample position.

For example, in the example of FIG. 7, the correction signal generation unit 154 refers to the coefficient table of FIG. 6, and the local maximum value counted by the sample number counting unit 152 based on the frame signal x ₀ [i] illustrated in FIG. The coefficient “0.5” corresponding to the number of samples between the extreme values from the minimum value to the minimum value or between the minimum value and the maximum value, for example, “4” is extracted.

Here, in the coefficient table of FIG. 6, the larger the number of samples, the smaller the coefficient value is for the following reason. That is, when the number of samples from an arbitrary extreme value to the next extreme value is large, the frequency of the frame signal x ₀ [i] is low, for example, filtering with a 22.1 kHz low pass filter (LPF). Is applied, harmonics of the low-frequency frame signal x ₀ [i] remain without being suppressed. Therefore, a sufficiently high sound quality can be maintained without adding a large high-frequency component, and the coefficient can be small.

On the other hand, when the number of samples from an arbitrary extreme value to the next extreme value is small, the frequency of the frame signal x ₀ [i] is high, for example, when filtering is performed by a low-pass filter of 22.1 kHz. The harmonics of the high-frequency frame signal x ₀ [i] are almost reduced. Therefore, the sound quality cannot be improved unless sufficient high-frequency components are added, so the coefficient needs to be large.

Subsequently, the correction signal generation unit 154 sets 0.5 extracted from the coefficient table to the difference value dl between the maximum value of the frame signal x ₀ [i] illustrated in FIG. The multiplied multiplication result Δdl is arranged at the sample position of the maximum value, and the multiplication result Δds obtained by multiplying the difference value ds between the minimum value of the frame signal and the sample value before one sampling by 0.5 is arranged at the sample position of the minimum value. Then, the correction signal co [i] shown in FIG. 7B is generated.

  Further, here, a correction signal for adding and subtracting the multiplication results Δdl and Δds to and from the sample position of the maximum value or the minimum value is generated. However, the sample position to be added or subtracted is not limited to this, for example, the maximum It is also possible to add / subtract to a predetermined number of sample positions before and after the value or minimum value.

For example, the correction signal generation unit 154 multiplies the difference value dl between the maximum value of the frame signal x ₀ [i] shown in FIG. 8A and the sample value before one sampling by 0.5 extracted from the coefficient table. The obtained multiplication result Δdl is arranged at one sample position before and after the maximum value, and the multiplication result Δds obtained by multiplying the difference value ds between the minimum value of the frame signal and the sample value before one sampling by 0.5 is set to 1 before and after the minimum value. The correction signal co [i] shown in FIG. 8B is generated at the sample positions. Further, the multiplication results Δdl and Δds are arranged at the sample positions of the local maximum value and the local minimum value and a predetermined number of sample positions before and after the local maximum value and the local minimum value, respectively, and FIG. 7B and FIG. It is also possible to generate a correction signal.

  In this way, the processing load is increased by a simple process of increasing the absolute value of the amplitude at an arbitrary sample position without complicated calculation for adding a specific high frequency component. It is possible to improve sound quality while reducing noise.

The delay unit 156 delays the frame signal x ₀ [i], which is the original signal, by the processing time in the extreme value specifying unit 150, the sample number counting unit 152, and the correction signal generating unit 154, as shown in FIG. As shown in FIG. 7B, FIG. 8A, and FIG. 8B, the frame signal x ₀ [i] and the correction signal co [i] are synchronized.

For example, the adder 158 adds the correction signal co [i] shown in FIGS. 7B and 8B to the frame signal x ₀ [i] shown in FIGS. 7A and 8A. The frame signal x ₀ ′ [i] subjected to the sound quality improvement processing as shown in FIG. 7C and FIG. In the present embodiment, the correction signal is added so as to be close to such a rectangular wave, thereby expanding the high frequency component and improving the sound quality.

However, running such a sound quality improvement process at random, since the frequency and the sampling frequency fs of the signal included in the frame signal x _{0 [i]} does not have a predetermined relationship, if the frame signal x ₀ Even if [i] is formed only from a regular sine wave, if the frequency is different from the sampling frequency fs, the number of samples from the minimum value to the maximum value differs from the number of samples from the maximum value to the minimum value. Therefore, the coefficient to be multiplied and the sample position to be added / subtracted differ every half cycle, and the sound quality improvement processing is biased. For example, when the number of samples in the entire period of the frame signal x ₀ [i] is “7”, the number of samples in one of the half periods is “4” and the other is “3”. The amount also depends on the number of samples.

In the present embodiment, as described above, the target of the sound quality improvement process is not the frame signal x ₀ [i], but the number of samples in the half cycle before and after the extreme value included in the frame signal x ₀ [i]. Since the fundamental wave signals b _k [i] are based on the same frequency, the sound improvement processing can be performed uniformly and uniformly.

For example, taking the first fundamental signal b _{k [i]} of one or more of the fundamental wave signal b _k to be inputted to the correction signal addition section 124 _[i] in the present embodiment example, extreme specifying unit described above The maximum and minimum values to be identified by 150 can be identified from the amplitude information of the sine wave component and cosine wave component of the fundamental wave signal b _k [i], and the sample positions of the maximum and minimum values are the fundamental wave. It can be identified from frequency information indicating the frequency of the signal b _k [i].

Also, the number of samples to be specified by the sample number counting unit 152 can be uniquely determined from the frequency information with reference to FIG. Therefore, the correction signal to be generated by the correction signal generation unit 154 can be uniquely derived from each information of the fundamental wave signal b _k [i].

As described above, the fundamental wave signal b _k [i] is formed with only a predetermined number of frequencies f _{k obtained} by dividing 1/2 of the sampling frequency fs by an integer. Accordingly, not only the number of samples is the same in the half cycle before and after the extreme value, but also the start and end points of the sine wave and cosine wave are located at the sample points. Then, the correction signal adding unit 124 can generate a correction signal by a simple process such as simply adding a uniform correction value.

FIG. 9 is an explanatory diagram for explaining the sound quality improvement processing in the correction signal adding unit 124 of the present embodiment. For example, in the sound quality improvement processing of the sine wave sin [i] shown in FIG. 9A, the maximum value and the minimum value appear for every sample number FS (here, 4) obtained from the frequency f _k . Even in the correction of the cosine wave cos [i] shown in FIG. 9B, the maximum value and the minimum value appear for each number of samples FS. Further, the sample position and coefficient to be added / subtracted are also determined according to the number of samples FS. Further, the value to be added or subtracted is uniquely determined according to the amplitude. Then, a correction signal for the sine wave sin [i] and the cosine wave cos [i] is uniquely derived from each information of the fundamental wave signal b _k [i]. Therefore, the correction signal adding unit 124 may change the frequency f _{k of} each fundamental wave signal, the amplitude of the sine wave sin [i], and the amplitude of the cosine wave cos [i] according to FIGS. ), The sound quality improvement process can be applied uniformly.

  Although the case where the sample number FS is an even number has been described here, a uniform correction signal can be generated in the same manner when the sample number FS is an odd number.

Further, since the value of the correction signal co [i] at each sample position of the sine wave and cosine wave is uniquely determined with respect to the frequency f _k of the fundamental wave signal b _k [i], the correction signal adding unit 124 A correction table in which the frequency f _{k of the} fundamental wave signal b _k [i] and the value of the correction signal co [i] at each sample position of the sine wave and cosine wave having an amplitude of 1 is created in advance It can also be left. Such a correction table may be held in a memory (not shown) or may be acquired from the communication network 106. Then, the correction signal adding unit 124 refers to the correction table, and at each sample position of the sine wave and cosine wave whose amplitude is 1 according to the frequency f _k of each of the one or more fundamental wave signals b _k [i]. The value of the correction signal co [i] is extracted, and the correction signal co [i] is generated by multiplying the amplitude of the sine wave and the amplitude of the cosine wave of each of the one or more fundamental wave signals b _k [i].

  Furthermore, since the sine wave or cosine wave included in the fundamental wave signal is proportional to the correction signal, it is possible to create a table by associating the sine wave or cosine wave with the signal obtained by adding the correction signal in advance. It is.

The processing load when generating the correction signal co [i] is significantly reduced by the configuration in which the correction signal co [i] is generated only for the fundamental wave signal b _k [i] excluding the residual signal d [i]. Therefore, the program can be simplified, and the cost can be reduced by using an inexpensive processing apparatus having a low processing capability.

Further, it becomes possible to add and subtract evenly the multiplication results obtained by multiplying the appropriate sample position by the appropriate coefficient with respect to all the fundamental wave signals b _k [i]. Since the same correction value proportional to the amplitude can be added to the same sample position of the wave signal b _k [i], it is possible to add a high-frequency signal without bias. As described above, the sound quality improvement can be made uniform by performing the sound improvement process uniformly on each signal included in the digital sound signal.

…（数式７） The residual signal adding unit 126 performs signal separation on one or more fundamental wave signals b _k [i] (frame signal x ₀ ′ [i]) _obtained by adding or subtracting the correction signal co [i] by the correction signal adding unit 124. The residual signal d [i] separated by the unit 122 is added to reconstruct the frame signal. Therefore, the reconstructed frame signal x ₀ ″ [i] is expressed by Equation 7. However, δs [i] and δc [i] in Equation 7 are respectively for a sine wave and a cosine wave having an amplitude of 1. The displacement amount is represented, i is 0 to L-1, k is 2, 3, 4,.

... (Formula 7)

  The overlap synthesizing unit 128 combines the frame signal reconstructed by the residual signal adding unit 126 and the previous frame signal (adjacent frames) so that a part of the digital audio signal overlaps. Combine and generate the final output signal.

  FIG. 10 is an explanatory diagram for explaining the operation of the overlap composition unit 128. The frame signal in FIG. 10 is a signal that is generated by the framing unit 120 and then passes through the signal separation unit 122, the correction signal addition unit 124, and the residual signal addition unit 126. The alphanumeric characters A, B, and C are , Corresponding to the digital audio signals A, B and C in FIG.

Specifically, the overlap synthesis unit 128 first performs the window function W shown in FIG. 10 on the reconstructed frame signal x ₀ ″ [i] (frame signal 0, frame signal 1, frame signal 2,...). When a window function using a sine wave window has already been applied in the framing unit 120, a window function using a sine wave window is also used in the overlap synthesis unit 128. In addition, a window function is used in the framing unit 120. If not, the Hanning window or the Blackman window is used, and the window function is not limited to such a case, and when two frame signals overlap, if the overlap part is synthesized and equal to the non-overlapping part Various existing window functions can be employed.

  When the frame signal 1 in FIG. 10 is input, the digital audio signal A of the frame signal 0 is already held, and the overlap synthesis unit 128 performs the digital audio signal A of the frame signal 0 and the rear signal of the frame signal 1. The digital audio signal A and the rear signal A ′ are added so that A ′ overlaps to generate a synthesized signal A ″. At the same time, the overlap synthesizing unit 128 converts the digital audio signal B of the frame signal 1 into the next time. When the frame signal 2 is input from the frequency time converter 146, the overlap synthesizer 128 receives the digital audio signal B of the frame signal 1 and the frame signal 1 as in the case of the frame signal 1. Then, the synthesized signal B ″ is generated by adding the rear signal B ′ of the frame signal 2 so as to overlap. The overlap synthesis unit 128 connects the synthesized signals A ″, B ″, C ″,... Generated in this way and outputs them as needed.

(Speech processing program)
Further, the above-described voice processing apparatus 100 can be realized using a computer.

  FIG. 11 is a functional block diagram showing a typical example of a computer (information processing apparatus) 200 that can analyze a digital audio signal and process the digital audio signal using the analysis result as the audio processing apparatus 100. It is. The computer 200 includes a central processing unit 210, a temporary storage device 212, an external storage device 214, an input unit 216, and an output unit 218.

  A central processing unit (CPU) 210 controls the entire computer 200 with programs and applications in the temporary storage device 212 and the external storage device 214. The temporary storage device 212 includes a RAM, an EEPROM, a nonvolatile RAM, and the like, and temporarily stores digital audio signals and the like processed by the central processing unit 210. The external storage device 214 includes a flash memory, an HDD, and the like, and stores a program processed by the central processing unit 210. The input unit 216 inputs a digital audio signal from the broadcast station 102 through broadcast waves, from the content server 104 through the communication network 106, or directly from the storage medium 108, and transmits it to the temporary storage device 212. The output unit 218 transfers the output signal generated by the computer 200 to the playback device 110.

  The sound quality improvement process described above is performed by the central processing unit 210 executing a program. Therefore, at the same time when the audio processing apparatus 100 is provided, the computer 200 performs frequency analysis of the digital audio signal, and removes the digital audio signal from one or more fundamental wave signals and one or more fundamental wave signals. A signal separation step that separates the signal into signals, a correction signal generation step that generates a correction signal that increases the absolute value of the amplitude for each of the one or more fundamental wave signals, and adds the correction signal to the fundamental wave signal; There is also provided an audio processing program for executing a residual signal adding step of adding a residual signal to one or a plurality of fundamental wave signals to which signals are added. In addition, this program may be read from a storage medium and loaded into a computer, or may be loaded into the computer 200 via the communication network 106.

(Audio processing method)
Next, a speech processing method for analyzing a digital speech signal using the speech processing apparatus 100 described above and processing the digital speech signal using the analysis result will be described.

  FIG. 12 is a flowchart showing the overall flow of the speech analysis / synthesis method. The framing unit 120 of the audio processing device 100 sequentially extracts the digital audio signal acquired by the audio processing device 100 in predetermined frame units (predetermined number of samples) to generate a frame signal (S300).

Subsequently, the signal separation unit 122 performs a frequency analysis of the frame signal based on the general harmonic analysis, and each of the predetermined number of fundamental wave signals b _k [i] having a predetermined number of different frequencies f _k is a frame signal. The difference signal e _k [i] is _obtained by subtracting from (S302). The signal separation unit 122 determines whether or not processing has been performed for all of the predetermined number of frequencies f _k (S304), and when not performed for all (NO in S304), repeats the differential signal derivation step S302.

Once accomplished for all frequencies _{f k} of a predetermined number (YES in S304), rearranges the nine frequencies _{f k} energy _{E k} is the ascending order of the difference signal (S306). Then, the signal separation unit 122 subtracts the fundamental wave signal b _k [i] for all the frequencies f _k or until the residual signal d [i] becomes equal to or lower than a predetermined energy (NO in S308). , the nine corresponding to the frequency _{f k} nine fundamental signal _b k [i], the order of the frequency _{f k} rearranged sequentially subtracted from the frame signal _x 0 [i], the residual signal d [i ] Is derived (S310). Thus, the signal separation unit 122 can separate the digital audio signal into one or a plurality of fundamental wave signals b _k [i] and a residual signal d [i].

  Then, the correction signal adding unit 124 generates a correction signal that increases the absolute value of the amplitude for each of the one or a plurality of fundamental wave signals, adds the correction signal to the fundamental wave signal (S312), and adds the residual signal. The unit 126 adds the residual signal to one or a plurality of fundamental signals to which the correction signals are added (S314).

  Finally, the overlap synthesizer 128 synthesizes the frame signal reconstructed by the residual signal adder 126 and the previous frame signal so as to partially overlap, and outputs the final output signal. Generate (S316).

  Even with the audio processing method described above, it is possible to uniformly improve the sound quality by uniformly performing audio improvement processing on each signal included in the digital audio signal.

  As mentioned above, although preferred embodiment of this invention was described referring an accompanying drawing, it cannot be overemphasized that this invention is not limited to this embodiment. It will be apparent to those skilled in the art that various changes and modifications can be made within the scope of the claims, and these are naturally within the technical scope of the present invention. Is done.

  Note that each step in the voice processing method of the present specification does not necessarily have to be processed in time series in the order described in the flowchart, and may include parallel or subroutine processing.

  The present invention can be used in an audio processing apparatus, an audio processing method, and an audio processing program that analyze a digital audio signal and process the digital audio signal using the analysis result.

DESCRIPTION OF SYMBOLS 100 ... Audio | voice processing apparatus 120 ... Framing part 122 ... Signal separation part 124 ... Correction signal addition part 126 ... Residual signal addition part 128 ... Overlap synthesis part 200 ... Computer

Claims (7)

A signal separation unit that performs frequency analysis of the input digital audio signal and separates the digital audio signal into one or more fundamental wave signals and a residual signal excluding the one or more fundamental wave signals;
A correction signal adding unit that generates a correction signal that increases an absolute value of an amplitude for each of the one or more fundamental wave signals and adds the correction signal to the fundamental wave signal;
A residual signal adding unit that adds the residual signal to the one or more fundamental wave signals to which the correction signals are added, and
With
The one or more fundamental wave signals are a plurality of fundamental wave signals having different frequencies,
The signal separation unit obtains a difference signal when a plurality of fundamental wave signals having the same frequency as the one or more fundamental wave signals are subtracted from the digital audio signal independently, and the difference signal has a lower energy in order An audio processing apparatus, wherein the digital audio signal is sequentially subtracted from the digital audio signal to separate the digital audio signal into one or more fundamental wave signals and the residual signal. A framing unit that cuts out a digital audio signal in units of a predetermined frame and generates a digital audio signal for each predetermined frame;
An overlap synthesizing unit that synthesizes the input digital audio signal in units of frames so that part of the digital audio signals of adjacent frames overlap;
Further comprising
The digital audio signal input to the signal separation unit is a digital audio signal divided into predetermined frames generated by the framing unit,
The audio processing apparatus according to claim 1, wherein the digital audio signal in units of frames input to the overlap synthesis unit is input from the residual signal addition unit.   The one or more fundamental wave signals are signals represented by a predetermined frequency and respective amplitudes of a sine wave and a cosine wave having the predetermined frequency. Voice processing device.   4. The audio according to claim 3, wherein the correction signal adding unit generates the correction signal according to a frequency of each of the one or more fundamental wave signals, an amplitude of a sine wave, and an amplitude of a cosine wave. Processing equipment. The correction signal adding unit is
Refer to the correction table in which the frequency of the fundamental wave signal and the value of the correction signal at each sample position of the sine wave and cosine wave having an amplitude of 1 are associated in advance,
The value of the correction signal at each sample position of the sine wave and cosine wave having the amplitude of 1 is extracted according to the frequency of each of the one or more fundamental wave signals, and the sine wave of each of the one or more fundamental wave signals The sound processing apparatus according to claim 4, wherein the correction signal is generated by multiplying the amplitude of the cosine wave and the amplitude of the cosine wave. Analyzing the frequency of the input digital audio signal, obtaining a differential signal when subtracting a plurality of fundamental signals having the same frequency as one or a plurality of fundamental signals having different frequencies from the digital audio signal, The one or more fundamental wave signals are sequentially subtracted from the digital audio signal in ascending order of energy of the difference signal to remove the one or more fundamental wave signals and the one or more fundamental wave signals from the digital audio signal. Separated into residual signals,
For each of the one or more fundamental wave signals, a correction signal that increases the absolute value of the amplitude is generated and added to the fundamental wave signal,
An audio processing method, wherein the residual signal is added to the one or more fundamental wave signals to which the correction signals are added. On the computer,
Analyzing the frequency of the input digital audio signal, obtaining a differential signal when subtracting a plurality of fundamental signals having the same frequency as one or a plurality of fundamental signals having different frequencies from the digital audio signal, The one or more fundamental wave signals are sequentially subtracted from the digital audio signal in ascending order of energy of the difference signal to remove the one or more fundamental wave signals and the one or more fundamental wave signals from the digital audio signal. A signal separation step for separating into a residual signal,
A correction signal generation step of generating a correction signal that increases an absolute value of an amplitude for each of the one or more fundamental wave signals and adding the correction signal to the fundamental wave signal;
A residual signal adding step of adding the residual signal to the one or more fundamental wave signals to which the correction signals are respectively added;
A voice processing program characterized by causing

Images