JP3510493B2 - Audio signal encoding / decoding method and recording medium recording the program - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an audio signal encoding / decoding method and a recording medium having a program recorded thereon.
More specifically, a voice (including music, background sound, computer-synthesized sound, etc.) signal is subjected to data compression coding into computer-processable code data, which is temporarily stored in the storage means, and when necessary, from the storage means. It is suitable to be applied to the purpose of reproducing the audio signal by subjecting the read code data to the data decompression / decoding process.

[0002]

2. Description of the Related Art Conventionally, when a voice signal is processed by a computer, 16-bit voice data is subjected to a logarithmic conversion method (A-1).
There are known methods such as conversion into 8 bits by aw, μ-law, etc., or compression into 4 bits by ADPCM (Adaptive Differential PCM). In recent years, in some baseball games and the like, live broadcasts of announcers are performed in accordance with the progress of the game as in the case of TV broadcasting, and the ADPCM method with a high data compression rate is adopted for recording and accumulating such voice. .

FIG. 8 is a diagram for explaining a conventional ADPCM system, which is APCM (Adaptive PCM) and DPCM (Differentientication).
al PCM) is shown in combination. Figure 8
(A) is a block diagram of a coding device, and such a device can be realized by a hardware configuration or a software configuration using a DSP. In the figure, the input audio signal x
(N) is subtracted from the predicted value x ″ (n) by the adder 11,
The difference signal d (n) is obtained. The predicted value x ″ (n) is calculated by the adaptive predictor 16 using the linear prediction method according to the following equation.

[0004]

[Equation 1] Here, a _i is a prediction coefficient, and the long-term autocorrelation function R _i (i = 0, 1, 2, ..., P) of the speech signal x ′ (n) is used to square the prediction error d (n). Is required to be minimized. At this time, the adder 15 is reproducing the audio signal x ′ (n) by the sum of the dequantized value of the quantized difference signal d ′ (n) and the predicted value x ″ (n), and the quantization error is It does not accumulate.

Further, the difference signal d (n) is quantized by the quantizer 12 with a step size Δ (n), and becomes a quantized difference signal d '(n). This step size Δ (n) is obtained by the following equation by multiplying the step size Δ (n−1) one sample before by the coefficient p by the step size adaptive controller 14.

[0006]

[Equation 2] Here, the coefficient p is determined by the code c (n-1) one sample before.

Further, the quantized difference signal d '(n) is the encoder 1
3 is coded to be coded data c (n), which is transmitted in real time to a communication partner or temporarily stored in a memory according to the purpose.

FIG. 8B is a block diagram of a decoding device, and such a device can be realized by a hardware structure or a software structure using a DSP. In the figure, the decoding side reproduces the difference signal d ′ (n) from the input coded data c (n) and the step size Δ (n) generated and managed by itself according to the same algorithm as the above-mentioned coding side. Finally, the predicted value x ″ (n) is added to the audio reproduction signal x ′ (n) to obtain the audio reproduction signal x ′ (n). n)
After being D / A converted to, it is amplified by the amplifier and added to the speaker. Therefore, a high data compression rate can be obtained, and a large amount of announcement information can be efficiently stored in the memory and reproduced.

[0009]

However, in recent years, this type of announcement information (generally including background sounds such as cheers) tends to increase more and more, which is good when a large capacity CD-ROM is used. If a ROM cartridge or the like having a limited capacity is used, it is necessary to further reduce the code data. In this case, if the sampling frequency of the audio signal is lowered, the code data can be reduced, but the sound quality is sacrificed.

It is therefore an object of the present invention to provide a voice signal coding / decoding method capable of further reducing code data without degrading sound quality, and a recording medium recording the program thereof.

[0011]

The above-mentioned problem is solved, for example, by referring to FIG.
It is solved by the method of. That is, according to the audio signal coding method of the present invention (1), the audio signal is divided into predetermined block lengths, each block signal is spectrum-analyzed to obtain a substantial frequency band for each block signal, Consecutively connect multiple block signals that can be regarded as having substantially the same frequency band in succession, sample audio signals at the minimum required sampling frequency for each group signal obtained, and Data compression encoding processing is performed.

By the way, in the long-term average spectral distribution of the speech waveform, the frequency components below 800 Hz are the majority (about 80%), and there is little difference between men and women except for the low range (below 200 Hz), and the language ( It is known that there is little difference between Japanese and English. However, when this is observed in a short-time range (several tens to several hundreds of ms), the spectrum distribution (occupied band) of the voice waveform changes variously.

Further, in a game machine or the like, the announcer does not always speak, and there is a gap between the announcement and the next announcement. Also, the conversation between the announcer and the baseball commentator is performed at a constant rhythm over a predetermined period, as in a normal conversation. Furthermore, the cheers of the audience and brass band music are often inserted between the announcements. At this time, in order to hear the brass band music with a desired sound quality, it is necessary to cover high frequency components. In this way, the short-time spectrum distribution (required band) of the voice (sound) waveform changes variously, and therefore,
There is room for further band compression.

Therefore, in the present invention (1), the input voice signal is divided by a predetermined block length L, the block signals are spectrally analyzed, and the substantial frequency bands k1, k1, k1, k1, k2, k2, k2, k
Find 1,…, etc. Here, k1 <k2. Further, from the first block, a plurality of block signals which can be regarded as consecutively substantially the same frequency band are sequentially concatenated to obtain a group signal {(k1, k1, k1, k1 group), (k2, k2 , K2 groups), ..., Etc.} and the minimum required sampling frequencies 2k1, 2k2 ,.
Etc., the audio signal is sampled, and the data compression encoding processing by ADPCM or the like is performed on the obtained entire sampling data series.

Therefore, a required band can be ensured for each group signal, so that high sound quality can always be maintained. On the other hand, since the number of sampling data is small in a place where the required bandwidth is low, the code data can be further reduced without impairing the sound quality. That is, the total data compression rate can be further increased.

Preferably, in the present invention (2), in the present invention (1), instead of the audio signal, audio data preliminarily sampled at a constant sampling frequency F _S (eg 40 KHz) is to be processed. None, and re-sampling of audio data is performed at the minimum necessary sampling frequency f _s (<F _S ) for each group data that is sequentially connected.

Therefore, even for voice (sound) data which has been digitized in advance, the code data can be further reduced without impairing the sound quality.

The audio signal decoding method of the present invention (3) is data compression-encoded by the method of the above invention (1).
Audio coded data that is required for each group signal.
Sampling frequency sampled at a low frequency
And each information of data size and code data
Is input and the audio code data for each group signal is sequentially subjected to data decompression decoding processing at each corresponding sampling cycle.

According to the present invention (3), the present invention (1)
It is voice code data that is data compression coded by the
The minimum frequency required for each group signal.
Sampling frequency, data size, and sign
Enter the data and each information,
The audio signal in each required band is properly reproduced by the configuration in which the audio code data for each audio signal is sequentially decompressed and decoded at each corresponding sampling period. In addition, such an audio signal decoding method can be combined with the encoding method of the present invention (1) or (2) to configure an optimal audio encoding system.

A computer-readable recording medium according to the present invention (4) is for causing a computer to execute the method of encoding / decoding an audio signal according to any one of the above-mentioned inventions (1) to (3). It is a record of the program.

[0021]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings. The same reference numerals denote the same or corresponding parts throughout the drawings.

FIG. 2 is a diagram showing the configuration of the sound data encoding device according to the embodiment. In FIG. 2, 1 is a CPU that performs main control / processing of the device, and 2 is a CPU that the CPU 1 executes, for example, as shown in FIGS. RAM, ROM, EEPR for storing sound data code processing program and necessary data
A main memory including an OM or the like, a display unit (DSP) including a CRT or a liquid crystal, a keyboard (KBD) for inputting processing information, a hard disk device (HDD) 5a as a secondary storage device, and 5a. Is a sound data file that stores unprocessed sound data, and 5b is a code data file that stores processed code data.

Although not shown, a reading device for reading sound data created externally in advance and a writing device for writing code data created internally to an external ROM cartridge 42 or CD-ROM 43 can be connected. Is. In addition, a sound signal in which voice, music, background sound, and the like are mixed may be simply referred to as an audio signal throughout this specification.

In the figure, a sound signal from the microphone 7 (or other recording device, sound synthesizing device, etc.) is sampled (A / D converted) by a clock signal SCK having a predetermined frequency F _s (for example, fixed at 40 KHz), and sound is generated. It is stored in the data file 5a.

The coding device divides the sound data into blocks of a predetermined length, and spectrally analyzes each block data to obtain a substantial frequency band for each block data.
It was obtained by sequentially connecting a plurality of block data that can be regarded as the substantially same frequency band continuously from the first block.
The sound data is resampled at the minimum required sampling frequency for each group data including one block data, and the obtained all sampling data series is subjected to data compression encoding processing by, for example, ADPCM, and this is a code data file. Store in 5b. This sound code data is stored in the game ROM cartridge 4
2 or written in the CD-ROM 43 together with 3DCG animation data and used as an information source of sound data.

3 and 4 are flowcharts (1) and (2) of the sound data coding process according to the embodiment, and FIGS. 5 and 6 are image diagrams (1) and (2) of the sound data coding process according to the embodiment. ), And the sound data encoding process according to the embodiment will be described in detail below with reference to these drawings.

[0027] In FIG. 3, to set the threshold t _h for determining the substantial presence / absence of the spectrum in step S1. Here, “substantially” means that the decoded sound is necessary to obtain a desired sound quality. In step S2, a block length L of sound data to be subjected to spectrum analysis (for example, L = 100 to 1000 words is selected in the range) is set. In step S3, a sound data file is opened and the data length of the sound data is l _en
To get.

FIG. 5A shows the storage structure of the sound data file. The sound data file includes a series of pre-sampled voices of the announcer's live broadcast corresponding to each scene in a baseball game and the like, together with the background sound (cheers, music, etc.) at a predetermined frequency F _s (for example, 40 KHz fixed). Sound data is recorded. The data length l _en differs for each scene (announcement content) of the game, and an EOF code is added after the last sound data. The information of the data length l _en is obtained in advance for easy processing and is written in a part of the sound data file.

Returning to FIG. 3, in step S4, the sound data is divided by the block length L, and m = [len /
L] +2, and m arrays D [m] are initialized with data “0”. Here, the operation [len / L] means rounding down after the decimal point. Further, +2 is a value in which m has a margin in order to cover a fractional block or the like.

FIG. 5B shows a mode of block division.
A series of sound data having a data length of l _en is divided by the block length L to be considered, and each is used as block data. In that case, the last fractional block is also treated as one block.

In the spectrum analysis described later, when each block data is cut out as it is (that is, multiplied by a rectangular window), a distorted spectrum is observed. Therefore, in practice, a Hamming window or the like is used to cut out in order to reduce this effect. Moreover, since the waveform is attenuated at both ends when the Hamming window is used, a method of moving the processing sections (each block) while partially overlapping them is adopted. However, for simplification of description, a method of cutting out the block data as it is will be described here.

Returning to FIG. 3, in step S5, a counter i for counting the block number is initialized to 0. In step S6, one block of sound data is read. In step S7, it is determined whether or not it is an EOF code.
If it is not the EOF code, it is determined in step S8 whether or not the sound data of the block is shorter than one block length L. If it is short, the short portion is filled with data "0" in step S9. If it is not short, the above step S9
Skip the process of. In step S10, one block of sound data is spectrally analyzed.

FIG. 5C shows sampling values s (0) to s (L-1) T of one block data. The sampling cycle T of one example is equivalent to T = 25 μs (F _s = 40 KHz), but since this encoding processing does not handle sampling data generated in real time, acquisition of each sampling value and its spectrum analysis are performed at high speed. can get. Here, the sound signal s (nT) consisting of discrete values is expressed by the following equation.

[0034]

[Equation 3] The discrete Fourier transform F (s) of the discrete signal s (nT) is expressed by the following equation.

[0035]

[Equation 4] If this is expressed by using the fundamental frequency f ₀ = 1 / LT as a parameter, the following equation is obtained.

[0036]

[Equation 5] Further, the calculation of this spectrum analysis is expressed in the form of a matrix as follows.

[0037]

[Equation 6] The actual calculation is FFT (Fast Fourier Transform)
To do faster.

FIG. 5D shows the block sound signal s (n
The spectral values F (0) to F ((L-1) f ₀ ) of T) are shown. This figure shows the case where the block length L = 100, and the fundamental frequency is f ₀ = 400 Hz. Further, in the case towards the lower frequencies from a high frequency in this example, the spectral values of F (kf ₀₎ exceeds the threshold value t _h.

[0039] Returning to FIG. 3, towards the lower frequency from In step S11 high frequencies, and detect the spectrum position k when the _{F (kf 0)> t h} . The spectral position k corresponds to the frequency of the substantially highest frequency component included in the block signal s (nT) (hereinafter, also referred to as a band). In step S12, the detected spectral position k is stored in the array D [i] (initially D [0]). In step S13, the block counter i is incremented by 1,
The process returns to step S6.

In step S6, the next one block data is read and the same processing as above is performed. In this way,
When the EOF code is detected in the determination in step S7, it means that the spectrum analysis for all blocks is completed. As a result, the process proceeds to step S14, and the contents of the counter i (the number of blocks analyzed by spectrum) at that time are saved in the register I. Then, the process proceeds to FIG.

In FIG. 4, in step S21, a counter i for counting the block number is initialized to 0.
In step S22, a counter j = for counting the number of blocks that can be regarded as having substantially the same band as the block on the front side.
Initialize to 1. In step S23, the contents (for example, k ₁ ) of the array D [i] (initially D [0]) is set in the register K. In step S24, + is added to the block counter i
After that, an array D [i] (first D
The contents of [1]) (for example, k ₁ ) are set. In step S25, it is determined whether K≈M.

There are several possible methods for determining whether or not K≈M. One is K ≈ M when strictly K = M
And the other one is K−α ≦ M ≦ K + α (however,
α is an integer such as 1, 2, ... By the way, in the former case, the number of consecutive blocks that are regarded as substantially the same band is reduced because they require exactly the same, and in the latter case, the substantially same band is regarded as the difference to some extent is regarded as the same. The number of consecutive blocks that are considered to be large increases.

In any case, if K≈M in the determination in step S25, the same block number counter j is incremented by 1 in step S26, and the process returns to step S24. Step S
At 24, the counter i is incremented by 1, and then the array D is placed in the register M.
The content (eg, k1) of [i] (this time D [2]) is read. In this way, a plurality of block strings that can be regarded as being substantially in the same band are detected, and eventually K is determined in the determination in step S25.
If not ≈M, the process proceeds to step S27.

FIG. 6A shows an example of extracting a resample block sequence (group data) considered to be in the same band. Here, for example, the bands for the first four blocks are k ₁ and
The bands for the next three blocks are k ₂ (> k ₁ ) and the next 4 bands, respectively.
The band for each block is determined to be k ₃ (> k ₂ ).

Returning to FIG. 4, in step S27, f _s = g
A new sampling frequency f _s is determined by the relationship of (K). Here, the sampling frequency f _s and the contents of the register K (initially k ₁ ) are related in advance by a function (or a table (not shown)) g, for example, f _s = 2
There is a relationship of k ₁ f ₀ . In step S28, f _c = h
The cutoff frequency f _c of the low-pass filter processing described later is determined based on the relationship (K). Here, the cutoff frequency f _c and the contents of the register K (initially k ₁ ) are related by a function (or a table (not shown)) h in advance, for example, f _c = k ₁ f _0. .

In step S29, sound data for j (initially 4) blocks is read. At this time, if necessary, j
Multiply the humming window corresponding to the sound data for blocks. In step S30, U and D are set to be relatively prime integers, and the minimum U and D that can be approximated to f _c ≈F _s · U / D are determined. For example, now the cutoff frequency f _c =
If it is 26.7 KHz, by setting U = 2 and D = 3, f _c ≈F _s · U / D = 40 KHz × 2/3 =
It satisfies 26.7 KHz. Then, in step S31, resampling processing of sound data (group data) for j blocks is performed by a known multi-rate filter processing.

In FIG. 6 (B) to FIG. 6 (D), the multi-rate
The image of filter processing is shown. In FIG. 6 (B),
First, the upsampler (U) inserts data "0" at the center of each sound data s (0) to s (L-1) T for four blocks. As a result, the signal band is doubled. In FIG. 6 (C), the the next cut-off frequency f _c by the low-pass filter (f _c), the filter sound data after the up-sampling (band-limited)
To process. As a result, new data is smoothly interpolated between the original four blocks of sound data s (0) to s (L-1) T. In FIG. 6 (D),
Next, the downsampler (D) downsamples (every two in this example) the sound data of the filter output. The sound data after resampling obtained in this manner is substantially equivalent to the sound data obtained by sampling the original sound signal of four block lengths from the beginning at f _s = 26.7 KHz (37.5 μs interval).

Returning to FIG. 4, in step S32, the resampled sound data is subjected to data compression encoding processing by, for example, the ADPCM system. The code processing by the ADPCM may be the same as that described with reference to FIG.
In step S33, the data size of j blocks (j ×
( Corresponding to L) and new sampling frequency fs information is output. In step S34, the above AD is continued.
The code data c (n) subjected to the PCM code processing is output. These pieces of information ( corresponding to j × L), fs, and their coded data c (n) are associated with each other and written in the coded data file 5b, and are used in the decoding process described later. In step S35, it is determined whether i <I (= maximum number of blocks). If i <I, the process returns to step S22, and the block indicated by the current counter i is regarded as the first block of the next group data, and the same process as above is repeated. Thus, when i <I no longer holds, the coding process for one sound data is ended.

FIG. 7 is a diagram showing the configuration of the game device according to the embodiment. Here, the configuration of the sound data decoding process according to the embodiment is provided.

In the figure, 30 is the main body of the game device, and 31 is
Is a CPU that performs main control of the game device (game program, reading control of 3DCG animation data and sound code data, reading control of game operation data by user, etc.) and game processing related to 3DCG animation (moving processing of game character, etc.) ( Game processor), 32 is RAM, ROM, E used by the CPU 31
A main memory (MM) including an EPROM, 33 is a 3D accelerator for perspectively converting input 3DCG animation data into a 2D screen, and 34 is a temporary storage of sound code data sequentially read from the ROM cartridge 42 or the CD-ROM 43. Buffer (BUF),
Reference numeral 35 denotes a peripheral interface (PIF) for accommodating an operation pad 41 described later in the game device, 36 denotes a ROM cartridge interface (ROM-CIF) capable of accommodating a ROM cartridge 42 detachably in the game device, and 37 denotes a CD-ROM 43 for the game. A CD-ROM driver 38 that can be detachably accommodated in the device and can be driven is provided by a public network (not shown) for this game device in order to download a game program online or to play a battle game with another game device. Communication control unit (COM) for connecting to
Is an ADP for decompressing and decoding sound code data
It is a CM decoding unit.

Further, reference numeral 41 denotes various control keys (start key,
An operation pad provided with position input means such as a mouse and a joystick, and a mask R for a game program (processing procedure, game parameters, 3DCG animation data, sound code data, etc.).
ROM cartridge (ROM-
C), 43 are CD-ROMs recording the same game programs as above, 44 is a communication line connected to the public network, 4
Reference numeral 5 is a display unit (DSP) made up of a CRT or liquid crystal for displaying a game, and 46 is a speaker for outputting a sound signal including voice.

In the ADPCM decoding unit 39, 25 is a waveform reproduction filter (for example, a cutoff frequency of 20 KHz).
A baseband processing unit (BBC) 26 including the above is a timing control unit that generates a processing timing signal SCK corresponding to the resampling frequency f _s of the sound data. Other configurations may be similar to those of the decoding device shown in FIG. Note that such an ADPCM decoding unit 39 can be realized by a hardware configuration or a software configuration using a DSP.

In the figure, the CPU 31 has a ROM cartridge 42 or a CD-ROM 43 as the game progresses.
The 3DCG animation data of the corresponding scene read out from the 3D accelerator 33 is supplied to the 3D accelerator 33, and the sound code data c (n) corresponding to the scene is appropriately transferred to the buffer 34. This sound code data c (n) is resampled for each group data by the sound data coding device shown in FIG.
It is a series of sound code data that is data compression coded by the PCM system. The timing control unit 26 generates a timing signal SCK having a corresponding frequency according to the information of the new sampling frequency f _s added to the sound code data c (n), and the sound code data c (n).
According to the information of the data size (corresponding to j × L) added to, the sound code data c (n) of the data size is sequentially read from the buffer 34.

In this way, the sound code data c (n) for the first j (= 4) blocks is subjected to data decompression decoding processing according to the timing signal SCK corresponding to the frequency 2k ₁ f ₀ . The timing signal SCK is also provided to the baseband processing unit (D / A conversion unit) 25, and the required band k ₁ f
A sound signal of ₀ is played. In the same manner, the sound code data c (n) for the next j (= 3) blocks is the timing signal S of the frequency 2k ₂ f ₀ (> 2k ₁ f ₀ ).
Data expansion decoding processing is performed according to CK, and the next j (= 4)
Sound code data c (n) for a block has a frequency of 2k
Data expansion decoding processing is performed according to the timing signal SCK of ₃ f ₀ (> 2k ₂ f ₀ ). In this way, a high-quality sound signal that is always maintained is applied to the speaker 46.

In the above embodiment, an example of application to the coding / decoding system by ADPCM is described, but the part related to the adaptive sampling rate conversion method according to the present invention is various other speech coding / decoding systems (waveforms). It is clear that it can be combined with a coding system, a source coding system, etc.).

Further, in the above embodiment, the voice data sampled at the predetermined frequency F _s (for example, 40 KHz) is input, and this is used as the minimum sampling frequency f for each group data having substantially the same required band. _s
Although the case where the resampling is performed is described above, the analog voice signal is directly input, and this is used as the minimum sampling frequency f for each group signal having substantially the same required band.
_s may be configured so as to sampling by is clear.

Further, in the above-mentioned embodiment, the block data which can be regarded as substantially the same band is connected to form the group data. However, without performing the group data, the re-sampling process and the code of the audio data are performed in the unit of the block data. / The device may be configured to perform the decoding process.

Although the preferred embodiments of the present invention have been described above, it goes without saying that various changes can be made to the configuration, control, processing, and combinations thereof of each part without departing from the spirit of the present invention. There is no.

[0059]

As described above, according to the present invention, the input voice signal is divided into blocks and each block signal is digitally processed at the minimum necessary sampling frequency. Therefore, high data quality can be achieved without sacrificing sound quality. A compression rate voice encoding / decoding process can be realized.

[Brief description of drawings]

FIG. 1 is a diagram illustrating the principle of the present invention.

FIG. 2 is a diagram showing a configuration of a sound data encoding device according to an embodiment.

FIG. 3 is a flowchart (1) of sound data coding processing according to the embodiment.

FIG. 4 is a flowchart (2) of sound data coding processing according to the embodiment.

FIG. 5 is an image diagram (1) of sound data coding processing according to the embodiment.

FIG. 6 is an image diagram (2) of sound data coding processing according to the embodiment.

FIG. 7 is a diagram showing a configuration of a game device according to an embodiment.

FIG. 8 is a diagram illustrating a conventional ADPCM system.

[Explanation of symbols]

1 CPU 2 main memory 3 Display (DSP) 4 keyboard (KBD) 5 Hard disk drive (HDD) 5a Sound data file 5b code data file 39 ADPCM Decoding Unit 42 ROM cartridge (ROM-C) 43 CD-ROM

─────────────────────────────────────────────────── ─── Continuation of the front page (58) Fields surveyed (Int.Cl. ⁷ , DB name) H03M 3/00-11/00 G10L 19/00

Claims (4)

(57) [Claims] 1. An audio signal is divided into a predetermined block length,
A group signal obtained by spectrally analyzing each block signal to obtain a substantial frequency band for each block signal, sequentially connecting a plurality of block signals that can be regarded as substantially the same frequency band consecutively from the first block An audio signal coding method characterized in that an audio signal is sampled at a required minimum sampling frequency for each time, and a data compression coding process is performed on the obtained all sampling data series. 2. Resampling of audio data at a minimum required sampling frequency for each group data that is sequentially processed, instead of processing the audio data, which is previously sampled at a constant sampling frequency. The method of encoding an audio signal according to claim 1, wherein 3. Data compression coding according to the method of claim 1.
Audio coded data that is required for each group signal
Sampling frequency sampled at the minimum frequency
It also includes information such as number, data size, and coded data.
Is input and the audio code data for each group signal is sequentially subjected to data decompression decoding processing at each corresponding sampling cycle. 4. A computer-readable recording medium having recorded therein a program for causing a computer to execute the method of encoding / decoding an audio signal according to claim 1. Description: