JP2000181497A - Device and method for reception and device method for communication - Google Patents

Abstract

PROBLEM TO BE SOLVED: To obtain a received voiced of improved auditory quality. SOLUTION: A postfilter 47 uses for processing a voice parameter code (e.g., α) based upon a transmitted signal sent from a transmitting device so as to generate a voice signal of 1st sampling frequency fs1 (8 KHz), but actually performs a post filter processing for a broadband voice signal of 2nd sampling frequency fs2 (16 KHz). For the purpose, the postfilter 47 performs the postfilter processing twice (=fs2/fs1) separately for 160 words each for the broadband voice signal of 320 samples (words) per frame (200 msec).

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a receiving apparatus and method for synthesizing an audio signal using an audio parameter code of an audio signal transmitted by communication or broadcasting, and a communication apparatus and method.

[0002]

2. Description of the Related Art In a conventional communication device, a sampling frequency of an input voice and an output voice on a receiving side are the same, and a voice frequency band is also the same. This is because the transmission band of the telephone line is narrow, for example, 300 to 3400 Hz, and the frequency band of the audio signal transmitted via the telephone line is limited.

[0003]

By the way, the sound quality of the sound output in the same sound frequency band as the input sound whose transmission band is limited is not so good. That is, the auditory quality is inferior. They also complain about the sound quality of digital mobile phones.

[0004] The present invention has been made in view of the above circumstances, and has as its object to provide a receiving apparatus and method, a communication apparatus, and a method capable of obtaining a received voice with improved auditory quality.

[0005]

In order to solve the above-mentioned problems, a receiving apparatus according to the present invention includes a transmitting signal transmitted from a transmitting apparatus to generate an audio signal having a first sampling frequency _fs1. a sampling rate converting means for converting the sampling frequency of the first band B ₁ of the audio signal to a second sampling frequency _{f s2 (f s2> f s1} ) generated using the speech parameters based code, the speech parameter codes and out-of-band component predicting unit to estimate the audio signal of the second sampling frequency f _s2 of the second band B ₂ is a band component of the first band B ₁ with the second at the sampling rate converting means Sampling frequency f _s2
A first audio signal having a bandwidth B ₁ which is a, and adding means for adding the second audio signal having a bandwidth B ₂ of the second sampling frequency f _s2, which was estimated by the out-of-band components estimating means, the addition Post-filter means for performing post-filter processing on the added output from the means.

Here, the post-filter means receives a decoded signal as input, and receives a spectrum shaping filter means for updating a filter coefficient in a first cycle, and an output from the spectrum shaping filter means, Gain adjusting means for updating the gain in a second cycle different from the first cycle. Further, the post filter means makes the second cycle longer than the first cycle.

[0007] In order to solve the above-mentioned problem, the receiving method according to the present invention generates the audio signal using the audio parameter code based on the transmission signal transmitted to generate the audio signal of the first sampling frequency f _s1 . first band B ₁ of the sampling frequency of the audio signal a second sampling frequency _{f s2 (f s2> f s1} ) is conversion obtained was converted output to the second
The first band B ₁ of the audio signal having the sampling frequency f _s2 of the first who guessed using the speech parameter codes
Second adding audio signal having the sampling frequency f _s2 of band B ₂ second is out-of-band component of the band B ₁ of performs post filtering processing on the addition output. In this method, the post-filter processing is performed on the sum output by f _s2 based on a voice parameter code based on a transmission signal transmitted from a transmission device to generate a voice signal of the first sampling frequency f _s1. / _Fs1 times.

[0008] In order to solve the above-mentioned problems, a communication apparatus according to the present invention comprises: a transmitting means for performing an encoding process on an input audio signal at a first sampling frequency f _s1 to generate a transmission signal; sampling frequency f sampling frequency of the first band B ₁ of the audio signal generated using the speech parameters code based on the transmission signal transmitted to produce a speech signal _s1 second sampling frequency f _s2 (f _s2> and converted output obtained by converting the f _s1), the second adder output the audio signal of the band B ₂ by adding of the second sampling frequency f _s2 of guessed using the speech parameter codes, post Receiving means for performing a filtering process.

Here, the receiving means is configured to generate a voice signal of a first band B ₁ using a voice parameter code based on a transmission signal transmitted to generate a voice signal of a first sampling frequency f _s1. a sampling rate converting means for converting the sampling frequency of the signal to a second sampling frequency _{f s 2 (f s2> f} s1), second a band component of the first band B ₁ with the speech parameter codes Out-of-band component estimating means for estimating the audio signal of the second sampling frequency f _s2 of the second band B ₂ , and the audio of the first band B _{1 set} to the second sampling frequency f _s2 by the sampling rate converting means signal and, a second adding means for adding the audio signal of the band B ₂ of the second sampling frequency f _s2, which was estimated by the out-of-band components estimating means, the addition output from the addition means And a post-filtering means for performing post-filtering.

The above-mentioned post-filter means receives a decoded signal, and receives the output from the spectrum-shaping filter means for updating the filter coefficient in the first cycle. First
And a gain adjusting means updated in a second cycle different from the cycle of Further, the post filter means makes the second cycle longer than the first cycle.

In order to solve the above-mentioned problems, the communication method according to the present invention performs an encoding process on an input audio signal at a first sampling frequency f _s1 to generate a transmission signal, and generates the transmission signal. sampling frequency f sampling frequency of the first band B ₁ of the audio signal generated using the speech parameters code based on the transmission signal transmitted to produce a _s1 second sampling frequency f _s2 (f
_s2 > f _s1 ) and the second sampling frequency f estimated using the above-mentioned speech parameter code.
a second adder output a sound signal having a bandwidth of B ₂ by adding of _s2, subjected to post-filtering.

[0012]

Embodiments of the present invention will be described below with reference to the drawings. This embodiment is a receiving apparatus 1 shown in FIG. 1, which is a specific example of a receiving apparatus according to the present invention. The receiving device 1 can be applied to a receiving side of a digital mobile phone, which is currently widely used as a personal digital cellular (PDC).

The receiving apparatus 1 obtains a second sampling frequency f _s2 (from a voice parameter code transmitted from a transmitting apparatus to be described later via a base station to generate a voice signal of a first sampling frequency f _s1 (f _s2 ). _fs2 > _fs1 ) is generated. 8 KHz is used as the first sampling frequency f _{s1 and} 16 KHz is used as the second sampling frequency f _s2 .

The voice parameter code received from the base station via the antenna 2 is stored in the memory 5a of the signal processing device 5 via the RF amplifier (RF receiving unit) 3 and the control unit 4.

The speech parameter code stored in the memory 5a of the signal processing device 5 is decoded by a decoding unit of the signal processing device 5, and then subjected to predetermined signal processing and output.

The output signal from the signal processing device 5 is D / A
After being converted into an analog signal by the converter 6, the analog signal is output from the speaker 10 via the anti-aliasing filter 7, the volume 8 and the amplifier 9. The control unit 4 is connected to, for example, a key operation unit 11 and an LCD display unit 12.

FIG. 2 shows a configuration of a transmitting device 15 that transmits the above-mentioned voice parameter code via, for example, a radio transmission path and a base station. This transmitting device 15 is also a PDC,
It can be applied to the transmitting side of a digital mobile phone which is widely used at present.

The audio signal input from the microphone 16 is stored in the memory 21a of the signal processing device 21 via the amplifier 17, the volume 18, the anti-aliasing filter 19 and the A / D converter 20.

The voice signal stored in the memory 21a is subjected to code processing in a voice coding unit inside the signal processing device 21, and is output as a voice parameter code. This voice parameter code is transmitted to the control unit 22 and the RF (RF transmission) amplifier 23
And transmitted to the base station via the antenna 24. The control unit 22 includes a key operation unit 25 and an LCD display unit 26.
Is connected.

Here, the speech coding unit inside the signal processing device 21 generates a speech parameter code in consideration of the narrow band limited by the radio transmission path. Generally, 300
A transmission band of 3 Hz to 3400 Hz is considered. The voice parameter code based on the transmission signal is supplied to the RF amplifier 23 via the control unit 22.

The speech parameter codes include a linear prediction (LPC) residual for the excitation source and a linear prediction coefficient α. Other examples include a lag LAG related to the pitch frequency and a frame power R0 in a frame of, for example, 20 msec.

The signal processing device 5 inside the receiving device 1 of FIG.
Comprises a decoder 27 shown in FIG. 3 and a bandwidth extension unit 32 shown in FIG.

The encoding method in the speech encoder in the signal processing device 21 of the transmitting device 15 shown in FIG.
If it is based on the CELP (Pitch Synchronous Innovation-CELP) coding method, the decoder 27 decodes the sound using the transmission signal based on the PSI-CELP coding, and outputs an output terminal 28. supplying an excitation source NExc _N decoded audio Snd _N, the linear prediction coefficient alpha _N to the output terminal 29, an output terminal 30 to. Here, the transmission signal by the PSI-CELP coding is one that has been transmitted to produce a speech signal of the first band B ₁ = ranging from 300 to 3400 Hz of the first sampling frequency f _s1 = 8 KHz.

[0024] The bandwidth extension unit 32 allows the decoder 27 to generate a speech signal of the first sampling frequency f _s1 (= 8 KHz) based on the PSI-CELP encoded transmission signal transmitted from the transmission device. The sampling frequency of the decoded audio Snd _N of the decoded _first band B ₁ (300 to 3400 Hz) is changed to the second sampling frequency f
_s2 (= 16 KHz), a sampling rate converter, and a linear prediction coefficient α _N obtained by the decoder 27 decoding the transmission signal by the PSI-CELP coding.
And a second band B ₂ (3400) of a second sampling frequency f _s2 (= 16 KHz) using the excitation source NExc _N and
Hz to 6000 Hz), and an audio signal of the first band B ₁ (300 to 3400 Hz) which is set to the second sampling frequency f _s2 (= 16 KHz) by the sampling rate converter. And a second band B ₂ (3400H) of the second sampling frequency f _s2 (= 16 KHz) estimated by the out-of-band component estimating means.
(z to 6000 Hz) audio signal;
The addition output from the addition means (300 Hz to 6000 H
z) is provided with post-filter means for performing post-filtering for spectral shaping and for improving the quality of hearing.

Here, the sampling rate conversion means is the up-sampling circuit 45 in FIG. Also,
The adding means is an adder 46, and the post filter means is a post filter 47. The out-of-band component estimating means includes an up-sampling circuit 45 in FIG.
, The adder 46 and the post filter 47.

Hereinafter, the configuration of the bandwidth extension unit 32 will be described in detail. First, the out-of-band component estimating means will be described. The out-of-band component estimating means includes a linear prediction coefficient → auto-correlation (α _N → r _N ) conversion circuit 36, an auto-correlation (r) broadband section 37, a wide band codebook (r _w CB) 38,
An autocorrelation → linear prediction coefficient (r _w → α _w ) conversion unit 39;
It comprises a PC synthesis section 40, an excitation source expansion section 41, a high-frequency extraction and suppression filter 42, and a multiplier 43.

The linear prediction coefficient α _N supplied from the input terminal 34 is supplied to a linear prediction coefficient → auto correlation (α _N → r _N ) conversion circuit 36. The α _N → r _N conversion circuit 36 converts the linear prediction coefficient α _N into an autocorrelation r _N and supplies it to the autocorrelation (r) widening unit 37. Autocorrelation (r) broadband unit 3
7 widens (extends) the autocorrelation r using a wideband codebook (r _w CB) 38. The wideband codebook (r _w CB) 38 is created in advance using the autocorrelation parameter r _w extracted from the wide band sound.

[0028] Using the wide band code book (r _w CB) 38, extended autocorrelation r _w autocorrelation (r) broadband portion 37 is expanded in the autocorrelation → linear prediction coefficients (r _w → α _w) conversion unit 39 Supplied. The r _w → α _w conversion unit 39 converts the extended auto-correlation r _w into the extended linear prediction coefficient α _w again, and supplies it to the LPC synthesis unit 40. LPC synthesis section 40 synthesizes a wideband speech based on the extended excitation source from the excitation source extension 41 to be described later with r _w → _{α w} wideband linear prediction coefficients from the conversion unit 39 alpha _w.

The extended excitation source from the excitation source extension unit 41 is also supplied to the LPC synthesis unit 40 as described above. Excitation source extension 41 extends the LPC residuals of the parameters relating to an excitation source supplied from an input terminal 35 (the LPC residuals referred to as the excitation source NExc _N.). FIG. 5 shows a detailed configuration of the excitation source extension unit 41.

First, the excitation source NExc _N supplied via the input terminal 35 is up-sampled by the up-sampling circuit 50. The output of the up-sampling circuit 50 is
The signal is sent from the output terminal 55 to the LPC synthesis section 40 via the LPF 51 and the boost circuit 52. That is, the excitation source NE
The signal obtained by up-sampling xc _N is used as the extended excitation source when synthesizing the audio signal. The boost circuit 52 boosts the extended excitation source when an affricate or a fricative is detected, and the boost amount is controlled by an output of the affricate detector 54. The affricate detection circuit 54 receives the autocorrelation r _N from the α _N → r _N conversion circuit 36 via the input terminal 53 and detects affricate and fricative noise.

The LPC synthesis section 40 synthesizes a wideband speech based on the wideband linear prediction coefficient α _w and the extended excitation source from the excitation source extension section 41. At this time, the LPC synthesis unit 40
Wideband linear prediction coefficient alpha _w, while updated every 2.5 msec (20 samples), synthesized wideband speech based on the extended excitation source. Generally, when the residual waveform is analyzed and synthesized by the harmonic encoding / decoding method, the envelope of the synthesized waveform becomes a very smooth and smooth waveform, and the LPC coefficient changes abruptly every 20 msec, resulting in abnormal noise. This is to prevent the occurrence of a problem. That is, if the LPC coefficient is gradually changed every 2.5 msec, generation of abnormal noise can be prevented.

The combined output of the LPC combining circuit 40 is supplied to a high-frequency extraction and suppression filter 42. The high-frequency extraction & suppression filter 42 removes a signal component in a frequency band of 300 Hz to 3400 Hz, and a second band B ₂ = 3400 Hz to 6
High frequency components are suppressed so as to extract a signal component of 000 Hz. The filter output from the filter 42 is multiplied by a multiplier 43 by a gain supplied from a terminal 44. The output (second band B ₂ = 3400 Hz to 6000 Hz) multiplied by the gain in the multiplier 43 is output to the adder 4
6.

Further, as described above, the bandwidth extension section 32 serves as the sampling rate conversion means as the input terminal 3.
3; first band B ₁ = 300-3400
Hz of the sampling frequency of the decoded audio Snd _N
An up-sampling circuit 45 for up-sampling from _s1 = 8 kHz to _fs2 = 16 kHz is provided.

Then, the up-sampling circuit 45 sets the sampling frequency to the second sampling frequency f _s2 = 16 k
Hz converted to the first band B ₁ = 300 Hz to 34
The adder 46 adds the 00 Hz audio signal component and the audio signal component of the second band B ₂ = 3400 Hz to 6000 Hz of the second sampling frequency f _s2 = 16 kHz, which is a multiplied output from the multiplier 43.

Further, a wideband audio signal Snd _w having a band of 300 to 6000 Hz and a sampling frequency of 16 kHz, which is an addition output from the adder 46, is supplied to the post filter 47.

[0036] The post-filter 47, the present applicant has already filed by disclosed in Japanese Patent Laid-Open No. 9-127996, is applied in the speech decoding method and apparatus technology, the wideband speech signal Snd _w Are subjected to spectral shaping and post-filter processing for improving the quality of hearing.

FIG. 6 shows a detailed configuration of the post filter 47. The spectrum shaping filter 131, which is a main part of the post filter 47, includes a formant emphasis filter 132 and a high-frequency emphasis filter 133. An output from the spectrum shaping filter 131 is a gain adjuster 134 for correcting a gain change due to spectrum shaping.
And the gain G of the gain adjuster 134
Is determined by the gain control circuit 136. The gain control circuit 136 compares the input and the output of the spectrum shaping filter 131 to calculate a gain change, and calculates a correction value of the gain G of the gain adjuster 134. Here, the input of the spectrum shaping filter 131 is the wideband audio signal Snd _w supplied via the terminal 135,
The output is connected to the post filter 4 via a terminal 137.
7 is a filter output derived from FIG.

In the bandwidth extension unit 32 having the above configuration,
The principle of operation will be described below. Bandwidth extension unit 32, 30 from the speech parameter codes to generate a first audio signal having a bandwidth B ₁ in 300Hz~3400Hz
A wideband speech coding parameter of 0 Hz to 6000 Hz is generated, and wideband LPC synthesis is performed. After that, the low frequency band (300 Hz to 3400 H
z) The side is replaced with the original sound upsampled to 16 KHz. That is, a high-pass filter is applied to keep only the high frequency band (3400 Hz to 6000 Hz), suppress high frequency components among these high frequency components, further adjust the gain, and then change the original sound (300 Hz to 3400 Hz).
To the up-sampled (second sampling frequency f _s2 ).

Here, to widen (or expand) the speech parameter code, it is necessary to broaden the linear prediction coefficient α _N and the excitation source NExc _N. Also,
In order to widen α _N , it is necessary to create a code book based on autocorrelation r, which is a parameter that can be mutually converted with α. The band of the autocorrelation r _N is widened by the quantization and inverse quantization by the code book.

First, the widening of the linear prediction coefficient α _N will be described. Paying attention to that α is a filter coefficient representing a spectrum envelope, it is once converted into an autocorrelation r _N which is a parameter representing another spectrum envelope which makes it easy to estimate the high-frequency side, and this is broadened. Extended)
Broadband from the autocorrelation r _w (or extension) linear prediction coefficients α _w
To the inverse. Vector quantization is used for extension. Narrowband autocorrelation r _n to vector quantization, may be obtained the corresponding r _w from that index.

Since a certain relationship is established between the narrowband autocorrelation and the wideband autocorrelation as described later, only a codebook based on the wideband autocorrelation needs to be prepared, and the narrowband autocorrelation can be vector-quantized by this. Wideband autocorrelation is obtained by inverse quantization.

Assuming that the narrow-band signal is obtained by band-limiting the wide-band signal, the wide-band auto-correlation and the narrow-band auto-correlation have a relationship expressed by the following equation (1).

[0043]

(Equation 1)

Here, φ is an autocorrelation, _xn is a narrow band signal, _xw is a wide band signal, and h is an impulse response of a band limiting filter.

Further, the following equation (2) is obtained from the relationship between the autocorrelation and the power spectrum.

[0046]

(Equation 2)

If another band-limiting filter having a frequency characteristic equal to the power characteristic of this band-limiting filter is considered, and this is set to H ′, the above equation (2) becomes the following equation (3)
It looks like an expression.

[0048]

(Equation 3)

The pass band and the stop band of this new filter are the same as those of the original band limiting filter, and the attenuation characteristic is squared. Therefore, this new filter can also be said to be a band limiting filter. With this in mind, narrowband autocorrelation is simplified to the convolution of broadband autocorrelation with the impulse response of a band-limited filter, ie, band-limited wideband autocorrelation. That is, the following equation (4) is obtained.

[0050]

(Equation 4)

As described above, when performing vector quantization of narrowband autocorrelation, if only a wideband codebook is prepared, a narrowband vector required at the time of quantization can be created by calculation. It turns out that there is no need to prepare a codebook.

[0052] Further, in order to have a curve r _w code vector of each wide-band autocorrelation increase or decrease monotonically decreasing or gradually, the H 'by no significant change be passed through a low-pass, r _n quantization directly It can be carried out in the r _w code book.
However, since the sampling frequency is 1/2, it is necessary to compare every other order.

The linear prediction coefficient α _N is extended because it can be extended more accurately by dividing it into voiced sound (V) and unvoiced sound (UV). Accordingly, two codebooks for V and UV are used.

Next, expansion of the excitation source will be described. P
In the SI-CELP, an excitation source in a narrow band is upsampled by inserting a zero value in the upsampling circuit 50 in FIG. 5 to generate aliasing distortion. Although this method is very simple, it can be said that the quality is sufficient as an excitation source because the difference between the power and the harmonic structure of the original voice is preserved.

Then, LPC synthesis is performed by the LPC synthesis circuit 40 using the wideband α _W obtained above and the wideband excitation source.

Since the quality of the speech synthesized by the wideband LPC is poor as it is, the original speech Snd _N of the codec output is replaced on the low frequency side. For this purpose, 3400 Hz or more is extracted from the synthesized sound, while the codec output is up-sampled to fs = 16 KHz, and these are added.

At this time, the gain by which the multiplier 43 multiplies the high frequency side can be adjusted by the gain adjuster according to the user's preference. This value is variable because individual differences between users are large. The value of the high-frequency gain is set in advance by an input from the user, and multiplication is performed with reference to this value.

Before addition, high-frequency extraction &
By applying a filtering that slightly suppresses a component of about 6 KHz or more by the suppression filter 42, a sound that is easy to hear is obtained. This filter coefficient is selectable, and processing is performed using a filter selected in advance, so that a higher frequency band can be selected as desired. The selection of this filter is also set by the user's input.

The processing using the filter 42 does not affect the power characteristics on the low frequency side, and may be performed after the addition. Alternatively, it is also possible to apply a filter that also affects the low-frequency side after addition. As described above, a wideband sound can be obtained.

Next, an operation in which the bandwidth extension unit 32 generates a wideband audio signal based on the above-described operation principle will be described with reference to a flowchart of FIG.

In step S1, the α _N → r _N conversion circuit 36 shown in FIG. 4 converts the linear prediction coefficient α _N decoded by the decoder 27 shown in FIG. 3 into an autocorrelation r _N. The audio signal Snd _N decoded by the decoder 27 is subjected to V / UV determination in step S2.

If the result of the determination in step S2 is V, in step S4 the voiced autocorrelation r _N is quantized. This quantization uses the narrowband V parameter obtained in step S3. That is, the parameters for the narrow band V obtained by comparing every other order from the code book 38 of the wide band V are used.

On the other hand, when the judgment result in step S2 is UV, in step S4 the autocorrelation r for unvoiced sound is quantized using the narrow band UV parameter obtained in step S3.

Then, in step S5, the wide band V
Inverse quantization using a codebook or a wideband UV codebook, which results in a wideband autocorrelation r _W. The broadband autocorrelation r _W is calculated in step S6 by using r _W → α _W conversion circuit 3
9 to α _W.

On the other hand, the excitation source from the decoder 27 is up-sampled by padding zeros between samples by the up-sampling circuit 50 shown in FIG. 5 in step S7, and widened by aliasing. This is supplied to the LPC synthesis circuit 40 as a broadband excitation source.

Then, in step S8, the LPC synthesis circuit 40 performs LPC synthesis on the wide band α _W and the wide band excitation source,
A wideband audio signal is obtained. Here, LPC synthesis is performed while updating the wideband linear prediction coefficient α _w every 2.5 msec (20 samples).

However, if this is the case, it is merely a wideband signal obtained by prediction, and the quality is poor because it contains errors due to prediction. In particular, regarding the frequency range of the input narrowband audio (300 Hz to 3400 Hz), it is better to use the original audio Snd _N (input audio) output from the codec as it is.

Therefore, of the synthesized sounds from the LPC synthesizing circuit 40, the frequency range of the input narrow-band sound is 300 to 34.
00 Hz in step S9 for high frequency extraction & suppression filter 42
By using a band stop filter (BSP).

Then, in step S10, the up-sampled original voice Snd _N by the up-sampling circuit 45 is added by the adder 46 in step S13. At this time, in step S11,
By filtering with a high-frequency extraction and suppression filter 42 that slightly suppresses components of about 6 KHz or more, the sound is easy to hear. This filter coefficient can be selected as described above.

Further, in step S12, the multiplier 43
To adjust the high-frequency gain according to the user's preference.

Here, the creation of a codebook used in the bandwidth extension unit 32 will be described. The creation of the codebook is generally well-known GLA (Generalized Llo
yd Algorithm). The wideband speech is divided into frames for a certain period of time, for example, every 20 msec, and the autocorrelation of a certain order, for example, the sixth order is obtained for each frame. Using the autocorrelation for each frame as training data, a six-dimensional codebook is created. At this time, the voiced sound and the unvoiced sound may be distinguished from each other, and the autocorrelation of the voiced sound and the autocorrelation of the unvoiced sound may be separately collected to create respective codebooks. In this case, the code book is referred to when α is expanded during the band expansion processing. At this time, a voiced sound or an unvoiced sound is determined, and the corresponding code book is used.

The bandwidth extension unit 32 uses a codebook for wideband voiced sound and a codebook for wideband unvoiced sound. The creation of the codebook for wideband voiced sounds will be described with reference to FIG. 8, and the creation of the codebook for wideband unvoiced sounds will be described with reference to FIG.

First, a wideband audio signal is prepared for learning,
In step S31 in FIG. 8, framing is performed for 20 msec per frame. Next, in step S32, for each frame, classification is performed as to whether it is a voiced sound (V) or an unvoiced sound (UV) by examining, for example, a frame energy, a value of zero crossing, and the like.

Then, in step S33, for example, the autocorrelation parameter r up to the sixth order is calculated in the wideband voiced sound frame. In step S34, for example, the autocorrelation parameter r up to the sixth order in the wideband unvoiced sound frame is obtained.

A wideband parameter is extracted from the sixth-order autocorrelation parameters of each frame in step S41 of FIG. 9 and a wideband V (UV) codebook of dimension 6 is created in step S42 by GLA.

As described above, the bandwidth extension unit 32 using the decoding method based on, for example, PSI-CELP can provide a high-quality wide-band audio signal whose sampling frequency has been converted from 8 KHz to 16 KHz.

Further, the bandwidth extending section 32 can perform post-filter processing on the wide-band audio signal for spectral shaping and improvement in audibility by the post-filter 47 whose configuration has been already described. The operation of the post filter 47 will be described in detail.

The characteristic PF (Z) of the spectrum shaping filter 131 shown in FIG. 6 can be expressed by the following equation (5) using the linear prediction coefficient αi.

[0079]

(Equation 5)

The fractional part of the equation (5) represents the formant enhancement filter characteristic, and the part (1-kz ^-1 ) represents the high-frequency enhancement filter characteristic. Β, γ, and k are constants, for example, β = 0.6, γ = 0.8, and k = 0.3.

The gain G of the gain adjustment circuit 134
Can be expressed as the following equation (6).

[0082]

(Equation 6)

In this equation, x (i) is the input of the spectrum shaping filter 131, that is, the wideband audio signal Snd.
_w and y (i) is the output of the spectral shaping filter.

Here, the spectrum shaping filter 13
As shown in FIG. 10, the update cycle of the coefficient of 1 is 20 samples and 2.5 msec, similarly to the update cycle of the coefficient α _w of the LPC synthesis unit 40. Update cycle is 160 samples, 2
It is 0 msec.

As described above, compared with the update cycle of the coefficient of the spectrum shaping filter 131 of the post filter 47,
By making the update cycle of the gain G of the gain adjustment circuit 134 long, an adverse effect due to the fluctuation of the gain adjustment is prevented.

That is, in a general post filter, the update cycle of the coefficient of the spectrum shaping filter and the update cycle of the gain are the same. At this time, if the update cycle of the gain is 20 samples and 2.5 msec, 1
As is clear from 0, the fluctuation within one pitch period causes click noise. Therefore, in the post filter 47, the gain switching cycle is set longer, for example, 160 samples for one frame, 20 ms.
By setting ec, it is possible to prevent a change in gain. Conversely, when the update cycle of the coefficient of the spectrum shaping filter 131 is increased to 160 samples and 20 msec, the post-filter characteristic cannot follow a short-time change in the audio spectrum, and good audibility quality cannot be improved. By shortening the update cycle of the filter coefficient to 20 samples and 2.5 msec, effective post-filter processing can be performed.

The post-filter 47 processes an audio parameter code (for example, α) based on a transmission signal transmitted from a transmission device to generate an audio signal of the first sampling frequency f _s1 (8 KHz). However, the post-filter processing is actually performed on the wideband audio signal Snd _{W having} the second sampling frequency f _s2 (16 KHz). For this reason,
The post-filter 47 performs the post-filter processing by the configuration shown in FIG. 6 on the wideband audio signal S of 320 samples (word) per frame (20 msec).
nd _W is applied twice (= _fs2 / _fs1 ) by 160 words.

By using such a post-filter 47, the bandwidth extending section 32 can effectively improve the spectral shaping and audible quality of a wideband audio signal. Therefore, the receiving device 1 including the bandwidth extension unit 32 as the signal processing device 5 can obtain a received voice with improved auditory quality.

Next, another specific example of the signal processing device 5 in the receiving device 1 of FIG. 1 will be described with reference to FIGS. Another specific example is the decoder 58 shown in FIG.
And a bandwidth extension unit 65 shown in FIG.

The encoding method in the speech encoder in the signal processing device 21 of the transmitting device 15 shown in FIG.
If it is determined to be based on the P (Vector Sum Excited Linear Prediction) coding method, the decoder 58 decodes the sound using the transmission signal by the VSELP coding, and outputs the decoded sound Snd _N to the output terminal 59. , The linear prediction coefficient α _N at the output terminal 60, the excitation source 1Exc _N1 at the output terminal 61, and the excitation source 2 at the output terminal 62.
Supply Exc _N2 .

The bandwidth extending section 65 has a configuration as shown in FIG. 12, and is different from the bandwidth extending section 32 shown in FIG. 4 in that an excitation source switching & extending section 68 is provided.

[0092] The PSI-CELP performs a process for allowing the codec itself, particularly the voiced sound V, to be heard in a perceptually smooth manner. Sounds like.
Therefore, when a broadband excitation source is created, a process as shown in FIG. 13 is performed by using an excitation source switching & extension unit 68 internally provided with a circuit for switching the excitation source. In the processing shown in FIG. 13, the excitation source processing shown in FIG.
To Step S89.

The excitation source of VSELP is represented by parameters β (long-term prediction coefficient), bL [i] (long-term filter state), γ (gain), and c1 [i] (excitation code vector) used for the codec.
β * bL [i] + γ * c1 [i], the former of which represents the pitch component and the latter of which represents the noise component, these are referred to as β * bL [i] and γ * c1 [i]. Divide, step S87
In a certain time range, if the former energy is large, it is considered that the voiced sound has a strong pitch. Therefore, the process proceeds to YES in step S88, the pulse train is used as the excitation source, and the process proceeds to NO in the portion without the pitch component to 0. Oppressed. If the energy is not large in step S87, the conventional narrow band excitation source is filled up with zero by PZ-CELP in step S89 by zero padding in step S89 to obtain a wide band excitation source. As a result, the auditory quality of voiced sound in VSELP is improved.

Then, in step S92, the original sound Snd _N is upsampled by the upsampling circuit 45 and added by the adder 46 in step S95. At this time, in step S91,
Filtering is performed by a high-frequency extraction and suppression filter 42 that slightly suppresses a component of about 6 KHz or more to make the sound easy to hear. This filter coefficient is selectable as described above.

Further, in step S93, the multiplier 43
To adjust the high-frequency gain according to the user's preference.

As described above, the bandwidth expansion unit 65 using the VSELP decoding method also sets the sampling frequency to 8 KH.
A high-quality wideband audio signal converted from z to 16 KHz can be provided.

Further, since the bandwidth extending section 65 has the same post-fill 47 as that shown in FIG. 6, the spectral shaping of the wideband audio signal and the quality of audibility can be effectively improved. Therefore, the bandwidth extension unit 65
Can improve the auditory quality.

As the signal processing device 5 inside the receiving device 1 shown in FIG. 1, a signal processing device comprising a bandwidth extending unit 70 shown in FIG. 14 and a decoding unit shown in FIG. It may be.

The decoding section shown in FIG.
It includes a decoder 77 and a PSI-CELP decoder 81, and switches the input of the transmission signal to the decoder 77 or 81 according to the encoding method of the transmission signal transmitted from the transmitting device side. That is, the transmission signal received via the input terminal 75 is switched by the changeover switch 76 in accordance with the type of the encoding method, that is, VSELP or PSI-CELP.

The two excitation sources 1Exc _N1 and 2Exc _N2 from the VSELP decoder 77 are supplied to the input terminals 66 and 67 of FIG. Moreover, the excitation source NExc _N from PSI-CELP decoder 81 input terminal 3 of FIG. 14 through the output terminal 82
5 is supplied.

The VSELP decoder 77 or PSI
-Linear prediction coefficient α _V or α _p from CELP decoder 81
Is selected by the changeover switch 80 in accordance with the type of the encoding method, and is supplied to the input terminal 34 of FIG.

Similarly, the VSELP decoder 77 or PS
The decoded audio from the I-CELP decoder 81 is also selected by the changeover switch 84 in accordance with the type of the above-mentioned encoding method, and is then supplied to the input terminal 33 of FIG.

Further, on the side of the bandwidth extension unit 70 shown in FIG. 14, the switching from the excitation source switching & extension unit 68 or the excitation source extension unit 41 is performed by the changeover switch 71 which switches according to the type of the encoding system. The source output is switched and supplied to the LPC synthesis unit 40.

Therefore, according to the bandwidth extension unit 70, it is possible to perform high-quality bandwidth extension by doubling the sampling frequency according to the type of the encoding method of the transmission signal transmitted from the transmitting device. And the provision of the post-filter 47, it is possible to effectively improve the spectral shaping and audible quality of the wideband audio signal. Therefore, the signal processing device 5 including the bandwidth extension unit 70 can improve the auditory quality.

Further, the signal processing device 5 inside the receiving device 1 shown in FIG. 1 may include a bandwidth extending unit 90 as shown in FIG.

The input terminal 91 of the bandwidth extension section 90 has L
An excitation source that is a PC residual is supplied. Also, the input terminal 9
2 is supplied with a linear prediction coefficient α. The excitation source from the input terminal 91 is sent to the LPC synthesis filter 93 and also to the up-sampling circuit 100. Input terminal 9
The linear prediction coefficient from 2 is sent to the LPC synthesis filter 93.

The LPC synthesis filter 93 has an input terminal 91
The speech signal is synthesized using the linear prediction coefficient from the input terminal 92 on the basis of the excitation source from. LPC synthesis filter 93
Is supplied to the up-sampling circuit 94.

The up-sampling circuit 94 up-samples the sampling frequency f _s1 of the audio signal synthesized by the LPC synthesis filter 93. The upsampled audio signal passes through only a predetermined band by a bandpass filter 95 and is supplied to an adder 96. This up-sampling circuit 94, band-pass filter 95, and adding circuit 96
Is a path for adding the signal of the component of the original frequency band to the synthesized audio signal.

Further, the linear prediction coefficient is sent from the LPC synthesis filter 93 to the linear prediction coefficient-autocorrelation conversion circuit 97. The linear prediction coefficient-autocorrelation conversion circuit 97 converts a linear prediction coefficient into an autocorrelation. This autocorrelation is sent to the narrow band codebook 98 and also to the affricate detection circuit 99.

The excitation source from the input terminal 91 is up-sampled by the up-sampling circuit 100 and is supplied to the LPC through the low-pass filter 101 and the boost circuit 102.
The signal is sent to the synthesis filter 103. Boost circuit 102
Is for boosting the excitation source when an affricate or fricative is detected, and the boost amount of the boost circuit 102 is controlled by the output of the affricate detector 99.

[0111] In the narrow band code book 98, autocorrelation information of narrow band audio signals obtained from a plurality of audio signal patterns is stored in advance as code vectors. In the narrowband codebook 98, the autocorrelation from the linear prediction coefficient-autocorrelation conversion circuit 97 is compared with the autocorrelation information stored in the narrowband codebook 98, and a matching process is performed. Then, the index of the best matching autocorrelation information is sent to wideband codebook 104.

In the wideband codebook 104, corresponding to the narrowband codebook 98, autocorrelation information of a wideband audio signal obtained from an audio signal of the same pattern as when the narrowband codebook 98 was created is used as a code vector. Is stored. When the best matching autocorrelation information is determined in the narrowband codebook 98, this index is sent to the wideband codebook 104, and the wideband codebook 104 determines the narrowband autocorrelation information determined to be the best match. Broadband autocorrelation information corresponding to the correlation information is read.

The wideband autocorrelation information read from the wideband codebook 104 is sent to an autocorrelation / linear prediction coefficient conversion circuit 105. The autocorrelation-linear prediction coefficient conversion circuit 105 converts the autocorrelation into a linear prediction coefficient. This linear prediction coefficient is the LPC synthesis filter 1
03 is sent.

The LPC synthesis filter 103 performs LPC synthesis, thereby synthesizing a wideband audio signal.
The audio signal synthesized by the LPC synthesis filter 103 is supplied to a high-frequency extraction and suppression filter 106 and a multiplier 107.

The high band extraction & suppression filter 106 is an LPC
A signal component in the frequency band of 300 Hz to 3400 Hz of the input narrow band audio signal is removed from the combined output from the combining filter 103 to extract a signal component of 3400 Hz or more, and suppresses a high frequency component according to the user's preference.
The multiplier 107 multiplies the filter output from the high-frequency extraction & suppression filter 106 by the gain adjusted from the terminal 108.

The adder 96 adds the original narrowband audio signal component via the BPF 95 to the multiplied output from the multiplier 107. Thereby, a wideband audio signal is obtained.

This audio signal is supplied to the post filter 109. The post filter 109 has the configuration shown in FIG. 6 and can effectively improve the spectral shaping and audibility of the wideband audio signal.

Therefore, even the receiving apparatus including the bandwidth extending unit 90 shown in FIG. 16 can generate a high-quality wide-band audio signal with a doubling of the sampling frequency, and can further improve the perceived quality.

Note that the signal processing device 5 inside the receiving device 1 is provided in each of the bandwidth extension units 32, 65, 70 and 90.
A noise reduction processing unit may be provided so as to be connected to the post-stage or the pre-stage of the post-filter.

This noise reduction processing section detects background noise by using a noise reduction processing method disclosed in Japanese Patent Application Laid-Open No. 7-193548, which has already been filed by the present applicant.
Oppress. This noise reduction processing method is based on a noise level of a background noise section detected from a voice parameter code based on a transmission signal transmitted from a transmission device to generate a voice signal of the first sampling frequency f _s1. A control signal is formed, and the content of the noise reduction processing is changed based on the control signal.

FIG. 17 shows a bandwidth expansion unit 32 in which a noise reduction processing unit 49 to which the above-described noise reduction processing method is applied is connected to the post-filter 47. Also, in FIG.
4 shows a detailed configuration of the noise reduction processing unit 49. Adder 4 above
6, which is an added output from 6, a band of 300 to 6000 Hz,
Broadband audio signal Sn with a sampling frequency of 16 kHz
d _w is supplied to the frame power calculation circuit 142 via the input terminal 141. Frame power calculation circuit 142
Calculates, for example, a square root of a root mean square, a so-called rms value, as the power for each frame having a period of 20 msec. The frame average power value calculated by the frame power calculation circuit 142 is supplied to the suppression ratio calculation circuit 143. The suppression ratio calculation circuit 143 calculates a suppression ratio, which is a coefficient for suppressing noise, using the frame average power calculated by the frame power calculation circuit 142. The suppression ratio calculated by the suppression ratio calculation circuit 143 is sent to the smoothing circuit 144. Smoothing circuit 14
4 performs a smoothing process on the suppression ratio calculated by the suppression ratio calculation circuit 143. The smoothing process is a process for avoiding discontinuity of connection between input audio signals divided in units of 160 samples in, for example, 20 msec. The suppression ratio that has been subjected to the smoothing processing is sent to the noise reduction circuit 145, where the noise reduction circuit 145 generates the wideband audio signal Sn.
Used to remove d _w noise.

The control signal obtained by discriminating the noise level detection signal input via the terminal 148 by the level discrimination circuit 147 is supplied to the suppression ratio calculation circuit 143.
In response to the control signal, for example, the threshold value of the above-described suppression ratio calculation is switched and controlled.

Next, the operation of the noise reduction processing section 49 will be described in detail. Frame power calculation circuit 1 of FIG.
42 is the wideband audio signal Sn per frame
Calculate the average power rms of d _w . This average power r
ms is supplied to the suppression ratio calculation circuit 143.

The suppression ratio calculation circuit 143 calculates the average power rm
s is compared with a certain threshold value nr1, and a suppression ratio scale is calculated based on the comparison result. That is, this suppression ratio s
cale is set to 1 when the average power rms is equal to or larger than the threshold value nr1, and is set to scale = rms / K (7) when the average power rms is smaller than the threshold value nr1. Here, K is a constant. In this case,
K = nr1.

Alternatively, the above equation (7) is calculated for all rms, and the suppression ratio scale
Is smaller than 1 (scale <1), the broadband audio signal Snd _w is multiplied by the suppression ratio scale calculated by the equation (7). This is the average power rms
In a frame where is smaller than the threshold value rn1, it means that the wideband audio signal Snd _w is multiplied by a gain smaller than 1. When the suppression ratio scale is 1 or more (scale ≧ 1) as a result of Expression (7), the wideband audio signal Snd _w is output without any processing. This is because, in a frame in which the suppression ratio scale is equal to the threshold value, the wideband audio signal Sn
This means multiplying d _w by a gain of 1. Therefore, by appropriately selecting the threshold value nr1, the gain is controlled to be small in a portion having a small power such as a noise portion, and the effect of substantially reducing noise is obtained. Note that the effect of noise suppression when using the above equation (7) is 倍 of the average power of the input signal.

Also, when noise suppression is too strong,
In the case of using in combination with a circuit for muting a certain level or lower, a second threshold value nr2 smaller than the above threshold value nr1 (this is set as a first threshold value) is set and input. The level is this second threshold value nr
In a region smaller than 2, it is preferable to reduce the suppression, that is, to weaken the strength of the expanding action of the expander.

By the way, since the input signal is not processed by discriminating between voice and noise, there is a tendency that the voice disappears when the power of voice such as consonant is relatively small. This phenomenon is particularly noticeable when a strong noise reduction is applied, and depending on the type of voice, a considerable sense of discomfort is felt. Therefore, it is necessary to consider how strong the noise reduction should be applied to the frame average power and from what size.

When the above-described processing is performed in units of frames, the connection of audio in frames becomes discontinuous.
When you hear it, you feel unnatural.

In consideration of the above, the above-mentioned suppression ratio scal
Set attack time and recovery time for e,
For example, by performing smoothing in frame units,
It is conceivable to prevent the unnatural feeling from appearing.

That is, as is clear from the configuration shown in FIG. 18, the suppression ratio scale calculated and calculated by the suppression ratio calculation circuit 143 is once subjected to a smoothing process by the smoothing circuit 144 and then to the noise reduction circuit 14.
I send it to 5.

The smoothing circuit 144 is provided to solve the above-described problem that occurs in the noise reduction processing, and sets the above-mentioned attack time and recovery time. In this example, the attack time is “0” and the recovery time is variable.

That is, when the calculated speech power of the current frame is larger than the previous frame, the value is used as it is. On the contrary, when the calculated speech power is smaller, smoothing is performed by a low-pass filter (LPF) having a predetermined characteristic to change the frame power. To avoid unnatural feeling of processing due to. The noise reduction circuit 145 supplies the wideband audio signal Snd _w to the suppression ratio sca via the smoothing circuit 144.
le to reduce the noise of the input signal Snd _w ,
An output signal with reduced noise is output from the output terminal 146.

The control signal obtained by discriminating the noise level detection signal via the terminal 148 by the level discrimination circuit 147 is supplied to the suppression ratio calculation circuit 143. In response to the control signal, the threshold for the above-described suppression ratio calculation is switched and controlled. That is, the threshold for the suppression ratio calculation is based on the noise level detection signal.

The noise level detection signal is a sound level of a background noise section detected from a sound parameter code based on a transmission signal transmitted from the transmitting device to generate a sound signal of the first sampling frequency f _s1. Can be represented by

Although not shown here, a noise section detection circuit for detecting a background noise section from the speech parameter code, and a noise level for detecting a noise level of the noise section detected by the noise section detection circuit are described. A detection circuit is required, and a noise level detection signal detected by the noise level detection circuit is supplied to a terminal 148.

Further, the noise reduction processing unit 49 uses a voice parameter code based on a transmission signal transmitted from the transmitting apparatus to generate a voice signal of the first sampling frequency f _s1 (8 KHz) for processing. However, the actual noise reduction processing is performed on the wideband audio signal Snd having the second sampling frequency f _s2 (16 KHz).
_W. For this reason, the noise reduction processing unit 49
The noise reduction processing by the configuration shown in FIG. 18 is 320 samples (words) per frame (20 msec).
Of the wide-band speech signal Snd _W, is subjected divided into 160 words min twice _{(= f s2 / f s1)} .

As described above, the noise reduction processing section 49
Since the noise component in the wideband audio signal can be reduced, the bandwidth extension unit 32 can effectively improve the spectrum shaping and audibility, and can output the wideband audio signal with the noise component reduced.

Note that the bandwidth extension units 32, 65, 70
Alternatively, the receiving device using the signal processing device provided with 90 may be integrated with the transmitting device to constitute a mobile phone device 110 as shown in FIG. This mobile phone device 110
Can also be applied to digital mobile phones that are currently being widely used as PDC.

In the portable telephone device 110, the audio signal input from the microphone 111 is stored in the memory 116a of the signal processing device 116 via the amplifier 112, the volume 113, the anti-aliasing filter 114, and the A / D converter 115. Is done.

The sound signal stored in the memory 116a is
The audio signal is encoded by an audio encoding unit in the signal processing device 116 and output as an audio parameter code.

This voice parameter code is transmitted to the control unit 117.
And RF (RF transmission) amplifier 118 and antenna 119
Is transmitted to the base station via.

Here, the audio encoding unit in the signal processing device 116 supplies an audio parameter code to the RF amplifier 118 via the control unit 117 in consideration of the narrow band limited by the transmission path.

The voice parameter code received from the base station via the antenna 119 is transmitted to the memory 12 of the signal processing unit 122 via the RF amplifier 118 and the control unit 117.
2a.

The voice parameter code stored in the memory 122a of the signal processing device 122 is decoded by the decoding unit of the signal processing device 122, subjected to predetermined signal processing, and output.

The output signal from the signal processing device 122 is D /
After being converted into an analog signal by the A converter 123, the analog signal is output from the speaker 127 via the anti-aliasing filter 124, the volume 125 and the amplifier 128.

Here, the signal processing device 122 includes the above-mentioned bandwidth extending unit 32, 65, 70 or 90. Therefore, the mobile phone device 110 shown in FIG. 19 can effectively improve the spectral shaping and audibility of a high-quality wideband audio signal whose sampling frequency is doubled on the receiving side, and can improve the noise component. Can be reduced.

In the above embodiment, the receiving device, the transmitting device, and the mobile phone device have been described as being applicable to a digital mobile phone device used as a PDC. However, a wideband (wideband) CDMA system, that is,
The present invention is also applicable to a mobile communication system having a wide frequency bandwidth.

[0148]

The receiving apparatus and the receiving method according to the present invention provide:
First sampling frequency f sampling frequency of the first band B ₁ of the audio signal generated using the speech parameters code based on the transmission signal transmitted to produce a speech signal _s1 second sampling frequency f _s2 ( _fs2 >
f to a second first band B ₁ of the audio signal having the sampling frequency f _s2 is converted obtained was converted output _s1), the sound parameter codes to use and guessed the first band of the band B ₁ Since the audio signal of the second sampling frequency f _s2 of the second band B ₂ , which is an external component, is added and the added output is subjected to post-filter processing, the spectral shaping of the wide band audio signal and the quality in audibility are effectively improved. Can be improved.

Further, the communication apparatus and the communication method according to the present invention perform an encoding process on the input audio signal at the first sampling frequency f _s1 to generate a transmission signal, and generate the transmission signal at the first sampling frequency f _s1 . converts the sampling frequency of the first band B ₁ of the audio signal generated using the speech parameters code based on the transmission signal transmitted to generate the second sampling frequency _{f s2 (f s2> f s1} ) a conversion output obtained, the second adder output the audio signal of the band B ₂ by adding of the second sampling frequency f _s2 of guessed using the speech parameter codes, so subjected to a post-fill process, the spectrum shaping In addition, it is possible to obtain a wideband audio signal whose audibility is effectively improved.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a configuration of a receiving device according to an embodiment of the present invention.

FIG. 2 is a block diagram showing a configuration of a transmitting apparatus for transmitting a voice parameter code to the receiving apparatus shown in FIG. 1 via a base station.

FIG. 3 is a diagram showing a PSI-CELP decoder which configures a signal processing device inside the receiving device shown in FIG. 1 together with a bandwidth extension unit.

FIG. 4 is a block diagram showing a bandwidth extension unit which configures the signal processing device inside the receiving device shown in FIG. 1 together with a PSI-CELP decoder.

FIG. 5 is a block diagram showing a detailed configuration of an excitation source extension unit included in the bandwidth extension unit shown in FIG. 4;

FIG. 6 is a block diagram illustrating a detailed configuration of a post filter included in the bandwidth extension unit illustrated in FIG. 4;

FIG. 7 is a flowchart illustrating a detailed operation of the bandwidth extension unit shown in FIG. 4;

FIG. 8 is a flowchart for explaining a training data generation process used for a codebook used in the bandwidth extension unit shown in FIG. 4;

FIG. 9 is a flowchart illustrating the generation of the code book.

FIG. 10 is a diagram for explaining a filter coefficient update cycle and a gain update cycle of the post filter.

FIG. 11 is a diagram showing a VSELP decoder included in another specific example of the signal processing device inside the receiving device shown in FIG. 1;

FIG. 12 is a block diagram showing a configuration of a bandwidth extension unit included in another specific example of the signal processing device inside the receiving device shown in FIG. 1;

FIG. 13 is a flowchart illustrating a detailed operation of the bandwidth extension unit shown in FIG. 12;

FIG. 14 is a block diagram showing a configuration of a bandwidth extension unit included in still another specific example of the signal processing device inside the receiving device shown in FIG. 1;

FIG. 15 is a block diagram showing a configuration of a decoding unit included in still another specific example of the signal processing device inside the receiving device shown in FIG. 1;

FIG. 16 is a block diagram showing a configuration of a bandwidth extending unit included in still another specific example of the signal processing device inside the receiving device shown in FIG. 1 above.

FIG. 17 is a block diagram showing a configuration in which a noise reduction processing unit is connected at a stage subsequent to the post filter in the bandwidth extension unit shown in FIG. 4;

FIG. 18 is a block diagram illustrating a detailed configuration of the noise reduction processing unit.

FIG. 19 includes a receiving device including a signal processing device using each of the above-described bandwidth extending units, integrated with a transmitting device.
FIG. 2 is a block diagram illustrating a configuration of a mobile phone device.

[Explanation of symbols]

1 receiving device, 15 transmitting device, 21 signal processing device,
27 PSI-CELP decoder, 32 bandwidth extension unit, 36 linear prediction coefficient → autocorrelation (α _N → r _N ) conversion circuit, 37 autocorrelation wideband unit, 38 wideband codebook, 39 autocorrelation → linear prediction coefficient conversion unit , 40 L
PC synthesis unit, 41 excitation source expansion unit, 47 post filter, 49 noise reduction processing unit

 ────────────────────────────────────────────────── ─── Continued on the front page (72) Inventor Shiro Omori 6-7-35 Kita-Shinagawa, Shinagawa-ku, Tokyo Sony Corporation F-term (reference) 5D045 CA01 CC02 CC07 5J064 AA01 BB10 BB12 BC01 BC07 BC08 BC12 BC18 BC19 BD02 9A001 EE05 GG11 HH15 KK60

Claims (27)

[Claims] 1. A sampling of the first first band B ₁ of the audio signal generated using the speech parameters code based on the transmission signal transmitted from the transmitting apparatus to generate an audio signal having the sampling frequency f _s1 Sampling rate conversion means for converting a frequency to a second sampling frequency f _s2 (f _s2 > f _s1 ); and a second band B which is an out-of-band component of the first band B ₁ using the voice parameter code. ₍₂₎ an out-of-band component estimating means for estimating an audio signal of a second sampling frequency f _s2, an audio signal of a first band B _{1 set to} a second sampling frequency f _s2 by the sampling rate converting means, adding means for adding second audio signal having a bandwidth B ₂ of the second sampling frequency f _s2, which was estimated by band components estimating means, Posutofu the addition output from said adding means Receiving device, characterized in that it comprises a post-filter means for performing filter processing. 2. The post-filtering means performs the post-filtering according to an audio parameter code based on a transmission signal transmitted from a transmitting device to generate an audio signal of the first sampling frequency f _s1. 2. The method according to claim 1, wherein the addition output is performed _fs2 / _fs1 times.
The receiving device according to the above. 3. A post-filter means, to which a decoded signal is input, and a spectrum shaping filter means for updating a filter coefficient in a first cycle; an output from the spectrum shaping filter means being input; 2. The receiving apparatus according to claim 1, further comprising: a gain adjusting unit that is updated in a second cycle different from the first cycle. 4. The receiving apparatus according to claim 3, wherein said post-filter means makes said second cycle longer than said first cycle. 5. The transmission signal is a PSI-CELP-coded or VSELP-coded signal, and the post-filter means is a PSI-CELP-coded or VSEL-coded signal.
Post-filter processing based on the speech parameter code obtained by decoding the P-coded signal
3. The receiving apparatus according to claim 2, wherein the reception is performed _s2 / _fs1 times. 6. The out-of-band component estimating means includes a part for extending the band of the linear prediction residual as the speech parameter code, and a part for extending the linear prediction coefficient as the speech parameter code to a wide band. The receiving device according to claim 1, wherein: 7. A wideband extension part of the linear prediction coefficient, wherein the first conversion part converts the linear prediction coefficient into an autocorrelation, and a code in which the autocorrelation of the first conversion part stores a wideband autocorrelation in advance. 7. An autocorrelation expansion unit that expands by referring to a book, and a second conversion unit that converts the expanded autocorrelation from the autocorrelation expansion unit into expanded linear prediction coefficients. Receiver. 8. The receiving apparatus according to claim 6, wherein the portion for band-expanding the linear prediction residual includes an up-sampling unit for up-sampling the linear prediction residual. 9. The transmission signal is a PSI-CELP coded or VSELP coded signal, and the out-of-band component estimating means is configured to perform the PSI-CELP coded or VSEL
Using a voice parameter code obtained by decoding the P-coded signal, a voice signal of a second sampling frequency f _s2 of a second band B ₂ which is an out-of-band component of the first band B ₁ is used. 2. The receiving device according to claim 1, wherein the receiving device estimates. 10. The receiving apparatus according to claim 1, further comprising a noise reduction processing means provided before or after said post filter means. 11. The noise reduction processing means according to claim 1, wherein said noise reduction processing means generates a speech signal of said first sampling frequency f _{s1 by converting} a background noise section detected from a speech parameter code based on a transmission signal transmitted from a transmission device. The receiving apparatus according to claim 10, wherein a control signal is formed according to a noise level, and the noise reduction processing is performed _fs2 / _fs1 times based on the control signal. 12. A sampling frequency of the first first band B ₁ of the audio signal generated using the speech parameters code based on the transmission signal transmitted to produce the audio signal having the sampling frequency f _s1 first the second sampling frequency f _s2 second first band B ₁ of the audio signal having the sampling frequency f _s2 is (f _s2> f _s1) conversion obtained was converted into an output, infer with the speech parameter codes Of the second band B ₂ , which is the out-of-band component of the first band B ₁
A sound signal having a sampling frequency of _fs2 , and post-filtering the added output. 13. The post-filter processing is performed on the sum output based on a voice parameter code based on a transmission signal transmitted from a transmission device to generate a voice signal of the first sampling frequency f _s1. 13. The receiving method according to claim 12, wherein the processing is performed _s2 / _fs1 times. 14. A transmitting means for performing an encoding process on an input audio signal at a first sampling frequency f _s1 to generate a transmission signal, and transmitting the audio signal at the first sampling frequency f _s1 to generate the audio signal. the sampling frequency of the first band B ₁ of the audio signal generated using the speech parameters code based on it has been transmitted signals a second sampling frequency f _s2 (f
_s2 > f _s1 ) and the second sampling frequency f estimated using the above-mentioned speech parameter code.
A communication device comprising: receiving means for performing post-filter processing on an addition output obtained by adding an audio signal of a _second band B2 of _s2 . 15. The receiving means comprises a first band B ₁ of the audio signal generated using the speech parameters code based on the transmission signal transmitted to produce a speech signal of a first sampling frequency f _s1 At the second sampling frequency f _s2 (f _s2 >
f _s1 ), and a voice signal having a second sampling frequency f _s2 of a second band B ₂ which is an out-of-band component of the first band B ₁ using the voice parameter code. and out-of-band component predicting unit to estimate a first audio signal having a bandwidth of B _1, which is the second sampling frequency f _s2 at the sampling rate converting means, a second sampling was estimated by the out-of-band components estimating means adding means for adding second audio signal having a bandwidth B ₂ of frequency f _s2, the communication of claim 14, characterized in that it comprises a post-filter means for performing post filtering processing to the addition output from said adding means apparatus. 16. The first filter according to claim 1, wherein
_Performing the post-filtering process f _s2 / f _s1 times on the added output according to a voice parameter code based on a transmission signal transmitted from the transmission device to generate a voice signal having a sampling frequency f _s1. The communication device according to claim 15, characterized in that: 17. The post-filter means, to which a decoded signal is input, a spectrum shaping filter means for updating a filter coefficient in a first cycle, an output from the spectrum shaping filter means, and a gain. 16. The communication apparatus according to claim 15, further comprising: a gain adjusting unit that is updated in a second cycle different from said first cycle. 18. The post filter according to claim 18, wherein
18. The communication device according to claim 17, wherein a period of the communication device is longer than the first period. 19. The transmission signal is a PSI-CELP-coded or VSELP-coded signal, and the post-filtering means is a PSI-CELP-coded or VSELP-coded signal.
_17. The communication apparatus according to claim 16, wherein post-filter processing based on a voice parameter code obtained by decoding the LP-coded signal is performed _fs2 / _{fs1 on the} added output. 20. The out-of-band component estimating means of the receiving means includes: a part for extending the band of the linear prediction residual as the speech parameter code; and a part for extending the linear prediction coefficient as the speech parameter code to a wide band. The communication device according to claim 15, comprising: 21. A part of the linear prediction coefficient extending to a wide band is a first conversion unit that converts the linear prediction coefficient into an autocorrelation, and a code in which the autocorrelation of the first conversion unit stores a wideband autocorrelation in advance. 21. The auto-correlation expansion unit that expands by referring to a book, and a second conversion unit that converts the expanded auto-correlation from the auto-correlation expansion unit into expanded linear prediction coefficients. Communication device. 22. The communication apparatus according to claim 20, wherein the portion that extends the band of the linear prediction residual includes an upsampler that upsamples the linear prediction residual. 23. The transmission signal is a PSI-CELP coded or VSELP coded signal, and the out-of-band component estimating means is configured to perform the PSI-CELP coded or VSELP coded signal.
An audio signal having a second sampling frequency f _s2 of a second band B ₂ , which is an out-of-band component of the first band B ₁ , is converted using an audio parameter code obtained by decoding the LP-encoded signal. The communication device according to claim 15, wherein the communication device estimates. 24. The communication apparatus according to claim 15, further comprising a noise reduction processing means provided before or after said post filter means. 25. The noise reduction processing means according to claim 1, wherein said noise reduction processing means generates a speech signal having the first sampling frequency f _s1 based on a speech parameter code based on a transmission signal transmitted from a transmission device. _25. The receiving apparatus according to claim 24, wherein a control signal is formed in accordance with the noise level, and the noise reduction processing is performed _fs2 / _fs1 times based on the control signal. And it generates a transmission signal by performing a coding process according to claim 26 first sampling frequency f _s1 for the input speech signal, the transmission signal transmitted in order to generate the first sampling frequency f _s1 the sampling frequency of the first band B ₁ of the audio signal generated using the speech parameters code based the second sampling frequency f _s2 (f _s2
> F _s1 ) and the second sampling frequency f _s2 estimated using the above speech parameter code.
And a second audio signal having a bandwidth B ₂ of the the sum output addition of,
A communication method comprising performing post-filter processing. 27. The noise reduction processing is performed on the added output by f _s2 / f based on a voice parameter code based on a transmission signal transmitted from a transmitting device to generate a voice signal of the first sampling frequency f _s1. The communication method according to claim 26, wherein _{fs1 is} performed once.