MXPA97005997A - Algebraic coding book with amplitude signal deimpulse selected for a rapidacodification of - Google Patents

Abstract

The present invention relates to a method for conducting a scan in a coding book, taking into consideration the coding of a sound signal, said coding book consisting of a group of combinations of amplitude / position pulses, each pulse combination amplitude / position defines the different positions L, and comprises both pulses, pulses of zero amplitude and non-zero amplitude pulses, assigned to the respective positions p = 1, 2 ... L of the combination, and assuming each pulse of non-zero amplitude, one of the possible amplitudes q, said method comprises the steps of: Pre-selecting from said coding book a sub-group of combinations of amplitude / position pulses in relation to the sound signal; said sub-group of combinations of amplitude / position pulses according to the coding of the sound signal, by means of which the complexity of the scanning is reduced The only way to explore a sub-group of amplitude / position pulse combinations of the coding book, where the pre-selection step comprises the previous establishment, of a function Sp of pre-assignment of the positions p = 1 , 2, ... L of valid amplitudes of said possible amplitudes q, in relation to the sound signal, and wherein the scanning step comprises only the exploration of the amplitude / position pulse combinations of said coding book which have non-zero amplitude impulses, which respect the previously established function

Description

ALGEBRAIC CODING BOOK WITH AMPLITUDE OF SELECTED SIGNAL IMPULSE FOR A QUICK VOICE CODING BACKGROUND OF THE INVENTION Field of the Invention The present invention relates to an improved technique for the digital coding of a sound signal, in particular, but not exclusively to a voice signal, in accordance with the transmission and synthesis of this sound signal. Brief Description of the Prior Art The demand for efficient techniques of digital voice coding with a good subjective quality / bit index interrelation is increasing for numerous applications such as voice transmission via satellite, land mobile transmission, digital radio or communication network , voice storage, voice response and wireless telephony. One of the best techniques of the previous art, which has the ability to achieve a good quality interrelation / bit index, is the so-called Linear Prediction of Excited Code (CELP) technique. According to this technique, the voice signal is displayed and processed in sample blocks L, (eg, vectors), where L is some predetermined number. The CELP technique makes use of a coding book. A coding book, in the context of CELP, is an indexed group of sequences of L-long samples, which we will refer to as L-dimensional code vectors (combinations of impulses that define the different positions of L, and comprise both pulses, zero amplitude pulse and non-zero amplitude pulse, assigned to the respective positions p = 1, 2, ... L of the combination). The coding book comprises, an index k that lies within a range of from 1 to M, where M represents the size of the coding book, which is sometimes expressed as a number of bits b: M «2b A coding book can be stored in a physical memory (for example, a look-up table), or we can refer to a mechanism to relate the index to a corresponding coding vector (for example, a formula). To synthesize the voice according to the CELP technique, each block of speech samples is synthesized by filtering the appropriate coding vector, from the coding book through time diversification filters, which model the characteristics of the signal spectrum voice. In the final encoder, the synthetic output energy is computed for all, or for a subgroup of the opposing encoding vectors of the coding book (scan in the coding book). The retained encoder vector is the one that produces the synthetic output energy, which is the closest to the original voice signal according to a perceptually weighted distortion measure.

The first class of coding books, are called "stochastic" coding books. One drawback of these coding books is that they often comprise substantial physical storage. They are "stochastic", for example, irregular in the sense that the trajectory from the index to the associated encoding vector includes query tables, which are the result of irregularly generated numbers, or of statistical techniques applied to groups of training for long dialogues. The size of the "stochastic" coding books tends to be limited by their storage and / or exploration complexity. A second class of coding books are the algebraic coding books. In contrast to "stochastic" coding books, algebraic coding books are not irregular, and do not require storage. An algebraic coding book, is a group of indexed encoding vectors, in which the amplitudes and positions of the pulses of the coding vector Ktn, can be derived from its index k, through a rule, which does not require, or it requires a minimum of physical storage. Therefore, the size of an algebraic book, is not limited by storage requirements. The algebraic coding books can also be designed for efficient exploration. OBJECTS OF THE INVENTION It is an object of the present invention, therefore, to provide a method and apparatus for drastically reducing the complexity of scanning a coding book by coding a sound signal, these being methods and apparatus applicable to an extensive class of coding books.

Another object of the present invention is to provide a method and apparatus having the ability to pre-select a subset of pulse combinations from the coding book, and to restrict the combinations that are explored by this subgroup, taking into account the reduction of the complexity of scanning the coding book. A further object of the present application is to increase the size of a coding book, allowing the individual nonzero amplitude impulses of the coding vectors, to assume at least one of the possible amplitudes q, without increasing the complexity of exploration.

SUMMARY OF THE INVENTION More particularly, there is provided a method for conducting a scan in a coding book, in accordance with the present invention, which takes into account the coding of a sound signal, this coding book consists of a group of combinations of amplitude / position pulses, each of the combinations of amplitude / position pulses defining the different positions L, and comprising both pulses, the amplitude zero pulses and the nonzero amplitude pulses assigned to the positions p = 1, 2, ... L of the combination, and assuming each non-zero amplitude pulse, one of the possible amplitudes q, the method comprising the steps of: pre-selecting from a coding book a subset of pulse combinations of amplitude / position, in relation to the sound signal; and exploring only the subset of amplitude / position pulse combinations, taking into consideration the coding of the sound signal, by which the exploration complexity is reduced, to explore only a subset of amplitude / position pulse combinations. The pre-selection step comprises pre-setting, in relation to the sound signal, a pre-assignment function Sp for the valid amplitudes of the positions p = 1, 2, ... L, of the possible amplitudes q , and the scanning step comprises, scanning only the combinations of amplitude / position pulses of the coding book, which have non-zero amplitude pulses with respect to the preset function. Still in accordance with the present invention, an apparatus for conducting a scan in a coding book is provided, taking into account the coding of a sound signal, this coding book consists of a group of amplitude / position pulse combinations, defining each combination of amplitude / position pulses the different positions of L, and comprising both pulses, the amplitude zero pulses and the non-zero amplitude pulses assigned to the respective positions p = 1, 2, ... L of the combination, and assuming each pulse of nonzero amplitude one of the possible amplitudes q, the apparatus comprising: means for pre-selecting from the coding book, a subset of combinations of amplitude / position pulses, in relation to the signal Sound; and means for scanning only the subset of the amplitude / position pulse combinations, taking into consideration the coding of the sound signal by means of which the complexity of the scan is reduced, thus being explored, only a subset of the amplitude / position pulse combinations of the coding book. The pre-selection means comprise means for presetting, in relation to the sound signal, a function Sp by pre-assigning valid amplitudes to the positions p = 1, 2, ... L of the possible amplitudes q, and the means Scanning means comprise means for limiting scanning to combinations of amplitude / position pulses from the encoding book, having non-zero amplitude pulses, with respect to the preset function. Furthermore still in accordance with the present invention, a cellular communication system is provided, to serve an extensive geographical area divided into a plurality of cells, which comprises: mobile transmitter / receiver units; cellular base stations, located respectively in said cells; means for controlling communication between cell-based stations; a bidirectional wireless communication sub-system, between each mobile unit located in a cell and the cellular base station of this cell, comprising this bidirectional wireless communication sub-system, both in the mobile unit and in the cellular base station ( a), a transmitter including means for encoding a speech signal, and means for transmitting the encoded speech signal, and (b) a receiver including means for receiving a transmitted coded speech signal, and means for decoding the speech signal. encoded voice received. The coding means of the speech signal, taking into account the coding of the speech signal, comprises an apparatus for conducting scanning in a coding book, this coding book consisting of a group of amplitude / position pulse combinations. , each combination of amplitude / position pulse defining the different positions L, and comprising both pulses, the amplitude zero pulses and the nocero amplitude pulses assigned to the respective positions p = 1, 2, ... L of the combination, and assuming each non-zero amplitude pulse, one of the possible amplitudes q, the apparatus comprising the conduction of the scan comprising: - means for pre-selecting, from the coding book, a subset of combinations of amplitude / position pulses, in relation to the speech signal; Y means to explore only the subgroup of the amplitude / position pulse combinations, by means of which the complexity of the scan is reduced, in the same way that only a subset of the amplitude / position pulse combinations of the book is explored. give coding; wherein the pre-selection means comprises means for pre-establishing, in relation to the sound signal, a function Sp by pre-assigning valid amplitudes to the positions p = 1, 2, ... L, among said possible q amplitudes, and wherein the scanning means comprises means for limiting the scanning of the amplitude / position pulse combinations of the coding book, having non-zero amplitude pulses that respect the preset function.

According to a preferred embodiment of the present invention, through means of the function Sp, one of the possible amplitudes q is pre-assigned as a valid amplitude for each position p, and the pre-established function is related when the pulses of non-zero amplitude of any combination of amplitude / position pulses, has an amplitude equal to the amplitude Sp pre-assigned to position p of the non-zero amplitude pulse. Preferably, the pre-allocation of one of the possible amplitudes for each position p, comprises the steps of: processing the sound signal, to produce a signal D of the emission surface filtered backward, and a residual signal R 'of tone-removed; the calculation of an estimate of the amplitude of the vector B, in response to the signal D of the emission surface filtered backward, and the residual signal R 'of tone-removed; and quantifying an estimate of the amplitude Bp of vector B, from each of the positions p, until obtaining the amplitude that will be selected for the position p. The calculation of the estimated amplitude of the vector B, advantageously comprises the step of adding the signal D of the emission surface filtered backwards, in standardized form: : 1-ß) up to the residual signal R 'of tone-removed in normalized form: «* 'L to obtain in this way, an estimate of the estimated amplitude of vector B of the form: wherein ß is a fixed constant, preferably having a value between 0 and 1. According to a preferred additional embodiment of the present invention, the quantization is performed on an estimate of the peak-normalized amplitude Bp of vector B, using the following expression: where the denominator max 1 B n • I n is a normalization factor, which represents a peak amplitude of the non-zero amplitude pulses. The pulse combinations may each comprise a number N of non-zero amplitude pulses, and the p positions of the non-zero amplitude pulses are advantageously restricted in accordance with at least one pulse permutation code. simple plugged-N. The scanning of the coding book preferably comprises maximizing a given ratio having a denominator a | < 2 computerized through means of closed circuits plugged-N, according to the following relationship: + C7 (pJ, p2) + 2 { / '(plfp2) * t7 (p ,, pj) + au'tp ^ pj) * 2í /' (p2, p3) + C / '(p .., p ,.) + 2ü' (p., p ..) + 2f '(p "P ..) + ... + 2C /' (p ... ,, p ..) where the computation for each closed circuit is written on a separate line from the outermost closed circuit to the innermost closed circuit of the closed circuits plugged N, where pn is the position of the non-zero amplitude pulse ntn of the combination, and where U '(px, Py) is a function dependent on the amplitude Spx pre-assigned to the position px, between the positions p and the amplitude Spy pre-assigned to the position p, between the positions p. In the above calculation, at least the innermost closed circuit of the closed circuits plugged in N, can be overlooked as long as the next inequality is true 5 D? rt'l p. p. < r ° " where Spn is the amplitude pre-assigned to position Pn, Dpn is the component Pntn of the emission surface of vector D, and Tp is a threshold related to vector D of the emission surface filtered backward. The objects, advantages and other features of the present invention, may be better appreciated by reading the following non-limiting description of a preferred embodiment thereof, provided by way of example only with reference to the drawings that accompany her.

BRIEF DESCRIPTION OF THE DRAWINGS In the attached drawings: Figure 1 is a schematic block diagram of a sound signal coding apparatus, comprising an amplitude selector and an optimization control, according to the present invention; Figure 2 is a schematic block diagram of a decoder apparatus associated with the coding apparatus of the Figure 1; Figure 3a is a sequence of the basic operations for the rapid scanning of the coding book according to the present invention, based on the selected signal pulse amplitudes; Figure 3b is a sequence of operations for pre-assigning one of the amplitudes q, up to each p-position of the amplitude / position pulse combinations; Figure 3c, is a sequence of the operations included in the N-plugged closed circuit exploration, in which the innermost closed circuit is ignored whenever the contribution of the first N-1 pulses is insufficient until it is determines the DAf numerator < T; Figure 4 is a schematic representation of the N-plugged closed circuits used in scanning the encoder book; and Figure 5 is a schematic block diagram illustrating the infrastructure of a typical cellular communication system.

DETAILED DESCRIPTION OF THE INVENTION Detailed Description of the Preferred Modality Figure 5 illustrates the infrastructure of a typical cellular communication system 1. Notwithstanding that the application of the method and the apparatus to conduct the exploration for a cellular communication system, in accordance with The present invention is described as a non-limiting example in the present specification, it should be understood that said method and apparatus can be used with the same advantages in many other types of communication systems, in which the coding of the sound signal. In a cellular communication system, such as system 1, a telecommunications service is provided in an extensive geographical area, by dividing that large area by a number of smaller cells. Each of the cells has a cellular base station 2, (Figure 5), to provide radio signal channels, and audio and data channels. The radio signal channels are used to send messages by radio mobile telephones (mobile transmission / reception units), such as 3, within the limits of the coverage area of the cellular base station (cell), and to issue calls to the other radio telephones, either inside or outside the cell of the base station, or in another communication network such as the Public Switched Telephone Network 4 (PSTN). Once the telephone radio 3 has a successful broadcast or reception of calls, an audio data channel is placed with the cellular base station 2, which corresponds to the cell in which the radio telephone 3 is placed, and the communication between the base station 2, and the telephone radio 3 occurs in that audio or data channel. The telephone radio 3 can also receive control or regulation information about the signal channel, while a call is in progress. If the telephone radio 3 leaves a cell during a call and another cell enters, the telephone radio conducts the call to an available audio or data channel in the new cell. In a similar way, if there is no call in progress, a control message is sent, in the signal channel, so that the telephone radio registers it in base station 2, associated with the new cell. In this way, it is possible that mobile communication covers an extensive geographical area. The cellular communication system 1, further comprises a terminal 5 for controlling the communication between the cellular base stations 2, and the Public Switched Telephone Network 4, for example, during communication between the telephone radio 3 and the PSTN 4, or between the radio telephone 3 in a first cell, and a radio telephone 3 in a second cell. Of course, a wireless two-way radio communication sub-system is required to establish communication between each radio telephone 3 placed in a cell, and the cellular base station 2 of that cell. Said wireless two-way radio communication system, normally comprises in both, the radio telephone 3 and the cellular base station 2, (a) a transmitter for encoding the voice signal, and for transmitting the encoded voice signal, through an antenna, such as 6 or 7, and (b), a receiver for receiving a coded voice signal transmission through the same antenna 6 or 7, and for decoding the received coded voice signal. As is known to those skilled in the art, voice coding is required in order to reduce the width of the band needed to transmit the voice through the wireless two-way radio communication system, for example, between a radio telephone 3 , and a base station 2. The intention of the present invention is to provide an efficient technique of digital voice coding, for a good interrelation between subjective quality / bit rate, for example, for bidirectional transmission of speech signals between a cellular base station 2, and a radio telephone 3, through an audio or data channel. Figure 1 is a schematic block diagram of a digital voice coding apparatus suitable for carrying out this efficient technique. The voice coding apparatus of Figure 1 is the same encoding apparatus as was illustrated in Figure 1 of the Paterna US Patent Application No. 97 / 927,528, in which an amplitude selector 1 12 has been added in accordance with the present invention. U.S. Patent Application No. 07 / 927,528, issued September 10, 1992, for an invention entitled "DYNAMIC CODING BOOK FOR EFFICIENT VOICE CODING, BASED ON ALGEBRAIC CODES".

The analog voice signal is sampled and processed in blocks. It should be understood that the present invention is not limited to the application to a voice signal. They can also be contemplated, encoders of other types of sound signal. In the illustrated example, the sampled speech input block S (Figure 1) comprises contive samples L. In the CELP literature, L is designated as the length of the "substructure", and is usually placed between 20 and 80. Also, we refer to the blocks of the samples L, as L-dimensional vectors. Several L-dimensional vectors are produced in the course of the coding process. Below is a list of those vectors, which appear in Figures 1 and 2, as well as a list of the transmitted parameters.

List of the main L-dimensional vectors S Vector of voice input; R 'Residual tone-removed vector; X Vector of the emission surface; D Vector of the emission surface filtered backwards; A | < Vector encoder of the index k of the book algebraic encoder; and Ck Innovation vector (filtered coding vector) List of transmitted parameters k index of the coding vector (feeding of the algebraic coding book); g Amplification; STP Short term prediction parameters (defining A (z)); and LTP Long-term prediction parameters (which define an amplification of the b-tone, and a delay of the T-tone).

Decoding Principles We believe that it is preferable to first describe the speech decoder apparatus of Figure 2, illustrating the different steps taken between the digital power supply (power supply "demultiplexer" 205), and the output power of the speech sample (output energy of the synthesis filter 204). The "demultiplexer" 205 extracts four different parameters of the received binary information from the digital power channel, called the k index, the g amplification, the short term prediction parameters STP, and long term prediction parameters LTP. The current of the L-dimensional vector S of the speech signal is synthesized based on these four parameters, as will be explained in the following description. The speech decoding apparatus of Figure 2 comprises a dynamic coding book 208, composed of an algebraic code generator 201 and an adaptive pre-filter 201, an amplifier 206, an aggregator 207, a long-term forecaster 203, and a synthesis filter 204, In a first step, the algebraic code generator 201, produces an encoding vector A | < in response to the index k. In a nd step, the encoding vector A | < it is processed by an adaptation prefilter 202 provided with the short term prediction parameters STP and / or the long term prediction parameters LTP, to produce an output energy innovation vector C ^ The purpose of the adaptation prefilter 202, is to dynamically control the frequency content of the output energy innovation vector Cfc, so that to improve the quality of the voice, for example, to reduce the audio distortion caused by the annoying frequencies for the human ear . The normal transfer functions F (z) are given below for the adaptation prefilter 202: Fa (z), is a "formant" prefilter in which 0 < Y-j < And 2 < 1, they are constant. This pre-filter improves the regions and formant jobs in a very effective way, especially in the coding index below 5 kbit / s. FD (z), is a tone filter where T is the variable time of the pitch delay, and bo is either constant or equal to the quantized long term tone prediction parameter, from the current or from the previous substructures . Ffc (z) is very effective in improving all the indices, the tone harmony frequencies. Therefore, F (z) normally includes a tone pre-filter, sometimes combined with a "formant" prefilter, that is: F U) - F. U) Fb (z) According to the CELP technique, the signal output energy for the voice recording S \ is obtained by the first scale of the innovation vector C | < , from the coding book 208 via the amplification g, through the amplifier 206. The aggregator 207, then adds the waveform to scale gC | < to the output energy E (the long-term prediction component of the signal excitation of the synthesis filter 204), of a long-term forecaster 203, supplied with the LTP parameters, placed in a closed feedback loop and having a transfer function b (z), defined as follows: B U). bz-t where b and T, respectively, are the amplification and pitch delay defined above. The forecaster 203 is a filter that has a transfer function, in accordance with the last received LTP b and T parameters, to periodically modulate the tone of the voice. This introduces the amplifications b and delays T of the appropriate samples. The signal of the component E + gCj < , constitutes the excitation of the filter signal d, e synthesis 204, which has a transfer function 1 / A (z) (where A (z) is defined in the description that follows). The filter 204 provides the correct formation of spectra according to the last received STP parameters. More specifically, the filter 204 models the resonant frequencies (formants). of the voice. The output power of the block S \ is the synthesized speech sample signal, which can be converted into an analogous signal with convenient "anti-aliasing" filtration, according to a technique well known in the art. There are many ways to design an algebraic code generator 201. An advanced method, described in the aforementioned US Patent Application No. 07 / 927,528, consists in the use of at least one N-plugged simple pulse permutation code. This concept will be illustrated, in the manner of a simple algebraic code generator 201. In this example, L = 40 and the group of 40-dimensional encoding vectors contain only pulses of nonzero amplitude N = 5, which will be called Sp- | , Sp2, Sp3, Sp4, Sp5. In this more complete observation, pj remains for the location of the impulse tp within the substructure (for example, p, has the range of from 0 to L-1). Suppose that, the Sp impulse? is forced to eight possible positions p-j, as shown below: P1 = 0.5, 10, 15.20, 25, 30, 35 = O + ßm -j; m < | = 0, 1 ... 7 Within these eight positions, which can be called "track" # 1, can they be freely swapped Sp? and seven pulses of zero amplitude. This is a "simple impulse permutation code". Now, we stratify five, said "simple impulse permutation codes" also restricting the positions of the residual impulses in a similar way (for example, track # 2, track # 3, track # 4 and track # 5) pi = 0, 5, 10, 15, 20, 25, 30, 35 = 0 + 8 mi p2 = 1, 6, 1 1, 16, 21, 26, 31, 36 = 1 +8 m2 p3 = 2, 7 , 12, 17, 22, 27, 32, 37 = 2 + 8 m3 p4 = 3, 8, 13, 18, 23, 28, 33, 38 = 3 + 8 M4 P5 = 4, 9, 14, 19, 24 , 29, 34, 39 = 4 + 8 M5 Note that the integers m, = 0, 1, ..., 7, fully define the position p, of each impulse Sp,. Therefore, a simple position kp index can be derived through the straight multiplexing of m, using the following relationship: k. - 4096 HV + 512 tr? 2 + 64 m3 + 8 4 + m « It should be noted that other coding books can be derived using the previous impulse tracks. For example, only 4 pulses can be used, where the first three pulses occupy the positions in the first three tracks, respectively, while the fourth pulse occupies, either the fourth or the fifth track with a bit, to specify which is track. This design gives rise to a book coding positions of 13 bits.

In the prior art, the non-zero amplitude pulses were assumed to have a fixed amplitude for all practical purposes, for reasons of the complexity of the vector vector scan. Undoubtedly, if the Spj impulse can assume one of the possible amplitudes q, it will be necessary to consider in the exploration, as many pulse amplitude combinations qN, as existing Spi impulses. For example, if the five impulses of the first impulse were allowed to take one of the possible amplitudes q = 4, for example Sp¡ = + 1, -1, +2, -2 instead of a fixed amplitude, the size of the algebraic coding book jumps from 15 to 15+ (5x2) bits = 25 bits; what it means, a thousand times more complex exploration. It is the purpose of the present invention to describe the surprising fact that a good performance can be achieved, with amplitude pulses q, without paying a very high price. The solution consists in limiting the exploration to a restricted subset of coding vectors. The method of selection of encoding vectors is related to the voice feeding, as will be described in the following description. The practical benefit of the present invention is to make possible an increase in the size of a dynamic algebraic coding book 208, allowing individual impulses to assume the different possible amplitudes, without increasing the complexity of scanning the coding book.

Principle of Coding The signal of the speech sample S, is coded on the bases of block by block, by means of the coding system of Figure 1, which is fragmented into 1 1 modules numbered from 102 to 112. The function and operation of most of those modules, remain unchanged, with respect to the description of the Paternal US Patent Application No. 07 / 927,528. Accordingly, notwithstanding that in the following description, the function and operation of each module will be explained at least briefly, it will concentrate on the matter that is new with respect to the description of the Paternal US Patent Application No. 07 / 927,528. For each block of samples L of the speech signal, a group of Linear Predictive Coding (LPC) parameters, called short term prediction (STP) parameters, is produced, according to the prior art technique, through the analyzer of spectrum 102 LPC. More specifically, the analyzer 102 models the spectrum characteristics of each block S of samples L. The feed block S of sample L is blanked by a bleach filter 103, which has the following function transfer, based on the current values of the STP parameters: M A U) ¿- ßi r- i i-0 where ao = 1, and z is the usual variable of the so-called z-transform. As illustrated in Figure 1, the bleach filter 103, produces a residual vector R.

A tone extractor 104 is used to compute and quantify the LTP parameters, called tone delay T and tone amplification g. The initial state of the extractor 104, is also placed for an Fs value, from an initial state extractor 1 10. In the Paternal US Patent Application No. 07 / 927,528, a detailed procedure for computerizing and quantifying the parameters is described. LTP, and we believe that it will be well known to those experts in the art. Accordingly, it will not be described further in the present description. A response characterizer of the filter 105, (Figure 1), is supplied with the STP and LTP parameters, to compute a response characterization of the FRC filter, to be used in the later steps. The FRC information consists of the following three components where n = 1, 2, ... L. * f (n): response of F (z) Note that F (z), usually includes the pre-filter tone. * h (n): response of 1 for f (n) where? P is a perceptual factor. More generally, h (n) is the impulse response of F (z) W (z) / A (z), which is the prefilter cascade F (z), the perceptual weighted filter W (z), and the synthesis filter 1 / A (z). Note that F (z) and 1 / A (z) are the same filters that were used in the decoder of Figure 2.

U (i, j): autocorrelation of h (n), according to the following expression: u (i, j) «? h (k-i + l) h (k-j + l) k = l for ls? L and ¡¡ysL; h (n) * 0 fbr n < l The long-term forecaster 106 is supplied with the past excitation signal (e.g., E + gCk of the previous substructure), to form the new component E, using the delay tone T, and of suitable amplification b. The initial state of the perceptual filter 107 is placed in the value Fs, supplied from the initial state of the extractor 1 10. The residual tone-removed vector R '= RE, which is calculated by a subtractor 121 (Figure 1) , then it is supplied to the perceptual filter 107, to obtain in the most recent filter output energy a vector of the emission surface X. As illustrated in Figure 1, the parameters STP, are applied to the filter 107, to vary its transfer function, in relation to those parameters. Basically, X »R '- P, where P represents the contribution of the long term prediction (LTP), includes the" resonance "of the past excitations. The MSE criterion, which applies to? now it can be established in the following annotations of the matrix. min | U | 2 * minfls' -s' l2 = minls' - l P-g? ^ H r] | s = mtnlX-g? kH r \ 2 where H, is a triangular-inferior Toeplitz matrix L x L, formed from the response h (n), as follows. The term h (0) occupies the diagonal of the matrix, and h (1), h (2), ... (L-1), occupies the respective lower diagonals. A backward filtration step is performed by the filter 108 of Figure 1. The zero position, with respect to the amplification outputs g, the derivative of the previous equation, until obtaining the optimal amplification, is as shown below: c dg X (? kH r) t 9 s \? kH rr |? 2 With this value for g, the minimization becomes.

The objective is to find the index k in particular, for which the minimization is achieved. Note that because || X || 2 is a fixed amount, the same index can be found by maximizing the following quantity: where D «(Xff) and < * * «| AtJf t | • In the backward filter 108, a vector of the filtered emission surface D = (XH) is computerized. The term "backward filtration" for this operation comes from the interpretation that (XH) is the reverse time filtering X. Only one amplitude selector 1 12 has been added to Figure 1 of the US Patent Application. Paterna No. 07 / 927,528 mentioned above. The function of the amplitude selector 1 12 is to move the encoder vectors A backwards, being explored by the optimization controller 10 &, for most of the most promising vector A encoders, to reduce in this way the complexity of the vector scan encoder As described in the foregoing description, each encoding vector Ak is a combination of amplitude / position pulses in wave form, which defines the different positions of L, p and comprises both pulses, amplitude zero pulses and amplitude pulses. non-zero, assigned to the positions p = 1, 2, ... L of the combination, respectively, where each pulse of non-zero amplitude assumes at least one of the different possible amplitudes q. Referring now to Figures 3a, 3b and 3c, the purpose of the amplitude selector '1 12, is to pre-establish a function Sp between the p positions of the waveform encoder vector, and the possible values of the amplitudes q impulse. The preset function Sp is derived in relation to the pre-scan speech signal from the encoding book. More specifically, pre-setting this function consists of pre-assigning, in relation to the voice signal, at least one of the possible amplitudes q, for each position p of the waveform (step 301 of Figure 3a). To preassign one of the amplitudes 1 to each position p of the waveform, an estimate of the amplitude of the vector B is calculated in response to the filtered surface vector of transmission backward D, and to the residual vector R 'of tone- removed. More specifically, the estimation of the amplitude of the vector B is calculated by the addition (sub-step 301 -1 of FIG. 3b) of the transmission surface vector filtered back D, in the normalized form: (l -ß > ÍD \ and residual residual-pitch vector R ', in the normalized form: 1 * 1 until obtaining in this way, an estimate of the amplitude of vector B of the form: D. _ R 1 ß = (l -ß) - ~ r + ß, where ß is a fixed constant, which has a normal value of 1/2 (the value of ß is chosen between 0 and 1, depending on the percentage of the non-zero amplitude pulses, used in the algebraic code). For each position p of the waveform, the amplitude Sp that is pre-assigned to that position p, is obtained by quantizing the amplitude estimate Bp of the corresponding vector B. More specifically, for each position p of the waveform, an estimate of the normalized peak amplitude Bp of vector B is quantized (sub-step 301-2 of Figure 3b), using the following expression: S = o (Bp / max | B) where Q (.) is the quantization function and is a normalization factor that represents a peak amplitude of the non-zero amplitude impulse. In the important special case, in which: - q = 2, that is that the impulse amplitudes can assume only two values (for example, Sp¡ = + -1); and - the density N / L of the non-zero amplitude pulse is less than or equal to 15%; the value of ß can be equal to zero; then the estimation of the amplitude of vector B is simply reduced to the vector of the emission surface filtered back D, and consequently Sp = signal (Dp) The purpose of the optimization controller 109, is to select the best coding vector Ak, from the algebraic coding book. The selection of the criterion is provided in the form of a proportion that is calculated for each encoding vector Ak, and maximized on all encoding vectors (step 303): where D = (XH) and aj = | A H t \ 2 • Because Ak, is an algebraic encoding vector that has nonzero amplitude pulse N, of respective Spj amplitudes, the numerator is the square of DA * r s D SB ¡TÍ P? P and the denominator is an energy term, which can be expressed as: where U (p, pp) is the correlation associated with two unit amplitude impulses, one in the position p and the other in the position pj This matrix is computed according to the previous equation in the characterizing filter of response 105, and included in order of parameters referred to as FRC, in the block diagram of Figure 1. A rapid method for computerizing this denominator (step 304), comprises the closed circuits plugged-N, which is illustrated in Figure 4, in which the linear notation S (i) and SS (l, j), is used in the place of the respective quantities "Sp¡" and "Sp¡ Spj." The computation of the denominator \ ^ It is the process that consumes most of the time The computations that contribute to which are made in each closed circuit of Figure 4, can be written in separate lines from the outermost closed circuit to the innermost closed circuit, such as follows: * sp 2ü (P., p2) * 2SPiSf > iU pl, p2) + sp_2v (P., i) + 2 [spsp u Pi, P3) + sp? S? > jU (P2lPi)) • * S (P «'+ 21S» V (P.'P »> * V *. ^ *»' + '' '+ SP..lV (P «-?' P ')] where p , is the position of the non-zero amplitude pulse itn.Note that the closed circuits plugged-N of Figure 4, make it possible to stop the non-zero amplitude pulses of the encoding vectors Ak, in accordance with the codes of simple impulse permutation plugged in. In the present invention, the complexity of the scan is drastically reduced, by restricting the subgroup of the coding vectors Ak, being scanned to the vectors encoding the non-zero amplitude pulses N, respect the preset function in step 301 of Figure 3a.The preset function is appreciated when the non-zero amplitude impulses N of an encoding vector Ak each have an amplitude equal to the amplitude pre-assigned to the position p of the non-zero amplitude impulse, this restriction of the subset of coding vectors, is performed by the first combination of the preset function Sp, with the inputs of the matrix U (i, j) (step 302 of Figure 3a) then, by using the closed circuits plugged-N of Figure 4, assuming that all the impulses S (i) are fixed, positive and with amplitude of unit (step 303). Therefore, even though the amplitude of non-zero pulses can take any of the possible q values in the algebraic coding book, the complexity of the scan is reduced for the case of fixed pulse amplitudes. More precisely, the matrix U (i, j), which is supplied by the response characterizing filter 105, is combined with the pre-established function according to the relationship that follows (step 302): o'i i. j) = st s, U (l, j) Where S, results from the selection method of amplitude selection 112, the one named S i is the amplitude selected for an individual position i, followed by the quantification of the estimation of the corresponding amplitude. With this new matrix, computerization for each closed circuit of the fast algorithm can be written on a separate line, from the outermost to the innermost circuit, as indicated below: < £ = v '(p tpx) * U' [p2, p2) * 2U, [pl, pi) * ü '(p ,, P3 > + 2U'ipl, pi) * 2U / (p2, p3). .. ... • • • • • • + u '(pH, pM) + 2ü'. { p tp ,,) + 2C /; (p., pw) + ... + 2u '(pN, pN) where px, is the position of the nonzero amplitude pulse xtn of the waveform, and where U '(P? .Py). is a function dependent on the pre-assigned Spx amplitude for a px position, between the p positions and the amplitude Spv pre-assigned for a position p, between the p positions. To further reduce the complexity of the scan, one can jump (cf Figure 3C) in particular, but not exclusively, the innermost closed loop provided the following inequality is true: t lf4l *. PaJ > Pa < *? where Spn is the amplitude pre-assigned to the position pn, Dpn is the pntn component of the transmission surface vector D, and TQ is a frequency related to the transmission surface vector filtered back D. The excitation signal global, E + gCk, is computerized through an aggregator 120 (Figure 1) from the signal gCk, from the controller 109 and the output energy E, from the predictor 106. The extractor module from the initial state 1 10, is constituted by a perceptual filter with a variable transfer function 1 / A (zP-1), in relation to the STP parameters, which subtracts from the residual signal R, the excitation signal E + gCk, for the sole purpose of obtaining the state of the final filter Fs, to be used as initial state in the filter 107 and in the tone extractor 104. The group of four parameters k, g, LTP and STP, are converted to the convenient digital channel format, by means of a "multiplexer" 1 1 1, which completes the procedure for encoding an S block of samples of the speech signal. Although the present invention has been described above, with reference to the preferred embodiments thereof, said embodiments may be modified without departing from the scope of the appended claims, and without departing from the spirit and nature of the present invention.

Claims (1)

1. Claims 1. A method for conducting a scan in a coding book, taking into consideration the coding of a sound signal, said coding book consists of a group of amplitude / position pulse combinations, each amplitude pulse combination / position defines the different positions L, and comprises both pulses, pulses of zero amplitude and non-zero amplitude pulses, assigned to the respective positions p = 1, 2, ... L of the combination, and assuming each pulse of non-zero amplitude, one of the possible amplitudes q, said method comprises the steps of: Pre-selecting from said coding book a sub-group of amplitude / position pulse combinations in relation to the sound signal; and Exploring only said sub-group of amplitude / position pulse combinations according to the coding of the sound signal, whereby the complexity of the scan is reduced to only the exploration of a 4 sub-group of combinations of amplitude / position pulses of the coding book; wherein the pre-selection step comprises the previous establishment of a function Sp of pre-assignment of the positions p = 1, 2, ... L of valid amplitudes of said possible amplitudes q, in relation to the sound signal , and wherein the scanning step comprises only the exploration of the amplitude / position pulse combinations of said coding book having non-zero amplitude pulses, which respect the previously established function. 2. The method as described in claim 1, further characterized in that the pre-establishment step comprises the pre-assignment, by means of the function Sp, of one of the possible amplitudes q, as the valid amplitude for each of the positions p, and where the pre-established function is respected when the non-zero amplitude pulses of a combination of amplitude / position pulses each have an amplitude equal to the amplitude pre-assigned by the function Sp to the position p of said non-zero amplitude pulse. The method as described in Claim 2, further characterized in that the step of pre-assigning one of the possible amplitudes q, to each position p, comprises the steps of: processing the sound signal to produce a emission signal surface D filtered backward and a residual signal R 'of tone removed; the calculation of an estimate of the vector B of amplitude in response to the filtered emission signal signal D backward and to the residual signal R 'of tone removed; and quantifying an amplitude estimate Bp of said vector B of each of said positions p, in order to obtain the amplitude to be selected for said position p. 4. The method as set forth in claim 3, further characterized in that the step of computing an estimated amplitude vector B comprises the step of adding the filtered emission signal D backwardly in a normalized manner: D U-β) Ibl to the residual signal R 'of the tone removed in the normalized form: 1 * 1 to obtain in this way, an estimated vector B of amplitude of the form: or "R 'ß = (l -ß) ~ * ß IDI |? '| where ß is a fixed constant. The method as described in Claim 4, further characterized in that β is a fixed constant having a value between 0 and 1. 6. The method as described in Claim 3, further characterized in that the quantization step for each of said positions p, comprises the quantization of a normalized peak amplitude estimate Bp of said vector B using the following expression: Bp / maxJfl where the denominator max 'B p I' n is a normalization factor that represents a peak amplitude of the non-zero amplitude pulses, 7. The method as described in Claim 1, further characterized in that said combinations of pulses each comprise a number N of non-zero amplitude pulses, said method further comprising the step of restricting the p positions of the non-zero amplitude pulses, according to at least a permutation code plugged-N of a single impulse. The method as described in Claim 3, further characterized in that each of said combinations of amplitude / position pulses comprises a number N of non-zero amplitude pulses, and wherein the scanning step comprises the step of maximization of a given proportion having a denominator ak2 computed by means of closed circuits plugged-N according to the following relation to; - u '(plfp) -?' ip ^ Pi) + 2U '(P1, P1) + C7' (p3, P3 > *? C / 'lPj.P,) * 2ü' (p2, P3) 2ul e., P + ... + ll < > --.- > * -) where the computation of each circuit is written on a separate line from the outermost circuit to the innermost circuit of the N-plugged closed circuits, where pn is the position of the nonzero amplitude pulse ntr1 of the combination, and in where U '(px, py) is a function that depends on the amplitude Spx, pre-assigned to the position px between the positions p and the amplitude Spy pre-assigned to the position p and between the positions p. 9. The method as described in claim 8, further characterized in that the step of maximizing said determined rate comprises the step of jumping at least the innermost circuit of the closed circuits plugged-N provided that the next inequality is true : where Spn is the amplitude pre-assigned to the position pn, Dpn is the component pn, h of the emission surface vector D and TD is a threshold related to the emission surface vector D filtered backward. 10. An apparatus for conducting a scan in a coding book in accordance with a sound coding signal, said coding book consisting of a group of amplitude / position pulse combinations, defining each of the different amplitude / position pulses positions L and comprising both pulses, the zero amplitude impulses and the non-zero amplitude pulses assigned to the respective positions p = 1, 2, ... L of the combination, and assuming each of the amplitude impulse no- zero one of the possible amplitudes q, said apparatus comprising: means for pre-selecting from said encoding book a sub-group of amplitude / position pulse combinations in relation to the sound signal; and means for scanning only said sub-group of amplitude / position pulse combinations according to coding of the sound signal, whereby the complexity of the scan is reduced to only the exploration of a sub-group of the combinations of amplitude / position impulses of the coding book. wherein the pre-selection means comprise means for presetting a function Sp by pre-assigning the positions p = 1, 2, ... L of valid amplitudes of said possible amplitudes q, in relation to the sound signal and wherein the Scanning means comprise means for limiting scanning to combinations of amplitude / position pulses of said encoding book having non-zero amplitude pulses, which respect the preset function. The apparatus as described in claim 10, further characterized in that the function of the pre-establishment means comprises means for pre-assigning, by means of the function Sp, one of the possible amplitudes q, such as valid amplitude for each of the p positions, and where the pre-established function is respected when the non-zero amplitude pulses of a combination of amplitude / position pulses have an amplitude each that is equal to the pre-established amplitude. assigned by the function Sp to the position p of said nonzero amplitude pulse. 12. The apparatus as described in Claim 1 1, further characterized in that the means for pre-allocating one of the possible amplitudes q to each of the positions p, comprises: means for processing a sound signal to produce a filtered surface emission signal D backward and a residual signal R 'of tone removed; means for calculating a vector B of amplitude estimated in response to the emission surface signal D filtered backward and to the residual signal R 'of tone removed; and means for quantifying, for each of said positions p, an amplitude estimate Bp of said vector B to obtain the amplitude to be selected for said position p. The apparatus as described in Claim 12, further characterized in that each of said means for calculating a vector B of estimated amplitude, comprises means for adding the emission surface signal D filtered backward in the normalized manner: (i 101 to the residual signal R 'of tone removed in the normalized form: R '1 * 1 to obtain in this way a vector B of estimated amplitude of the form: ß «d -ß) ~ + ß - ~« where ß is a fixed constant. 14. The apparatus as described in Claim 13, further characterized in that β is a fixed constant having a value between 0 and 1. 15. The apparatus as described in Claim 12, further characterized by each of the means of quantization comprises means for quantifying, a Bp estimate of normalized peak amplitude of said vector B for each of said positions p, using the following expression Vma? LflJ where the denominator max fl is a normalization factor that represents a peak amplitude of the non-zero amplitude pulses. The apparatus as described in Claim 10, further characterized in that said pulse combinations each comprise a number N of non-zero amplitude pulses, said apparatus further comprising means for restricting the p positions of the amplitude pulses. non-zero according to at least one plug-in permutation coding-N of a single pulse. The apparatus as described in Claim 12, further characterized in that said combinations of amplitude / position pulses each comprise a number N of nonzero amplitude pulses, and wherein the scanning means comprise means for maximizing a given ratio having a denominator otk and means for computing said denominator to 2 by means of closed circuits plugged-N according to the following relation: a2k - O ^ P ^ P * U '(P1, P2) * 2 V , (p, pl) +? ' i, Pi) * 2V '(P?> P¿ * 2 t7' ^ í 'P3 where the computation of each circuit is written on a separate line from the outermost circuit to a more inner circuit of the closed circuits plugged-N N, where pn is the position of the nonzero amplitude pulse ntn of the combination and in where U '(px, py) is a function dependent on the amplitude Spx pre-assigned to the position px between the positions p and the amplitude Spy pre-assigned to the position p and between the positions p. The apparatus as described in Claim 17, further characterized in that said means for calculating the denominator ak2 comprises means for jumping at least the innermost circuit of the closed circuits plugged-N provided that the following inequality is true: , where Spn is the amplitude pre-assigned to position pn, Dpn is the pntn component of vector D of the emission surface, and T ^ is a threshold related to vector D of the emission surface filtered backward. 19. A cellular communication system for serving an extensive geographic area, divided into a plurality of cells, which comprises: mobile transmitter / receiver units; cellular base stations, located respectively in said cells; means for controlling communication between cell-based stations; a wireless bidirectional communication sub-system, between each mobile unit located in a cell and the cellular base station of said cell, said wireless bidirectional communication sub-system in both mobile units and the cellular station base: (a ) a transmitter including means for encoding a speech signal, and means for transmitting the encoded speech signal, and (b) a receiver including means for receiving a transmitted coded speech signal, and means for decoding the received coded speech signal; wherein said means for encoding the speech signal comprises, an apparatus for conducting a scan in an encoder book, taking into consideration the encoding of the speech signal, said encoding book consisting of a group of amplitude / position pulse combinations. , each combination of amplitude / position pulses defining the different positions of L and comprising both pulses, the zero amplitude pulses and the non-zero amplitude pulses assigned to the respective positions p-1, 2 L of the combination, and assuming each pulse of zero amplitude, one of the possible amplitudes q, said apparatus comprising to drive the scanning: means for pre-selecting from said coding book, a subset of amplitude / position pulse combinations in relation to the speech signal; and means for scanning only said subset of amplitude / position pulse combinations, taking into consideration the coding of the speech signal, so that the complexity of the scan is reduced only to the exploration of a subset of amplitude pulse combinations. /position; where the pre-selection means comprise means pair «| pre-establish, in relation to the sound signal, a function Sp pre-assigned to the valid amplitudes of the positions p = 1, 2, ... L, of between said possible amplitudes q, and where the scanning means they comprise means for limiting the scanning to the combinations of amplitude / position pulses of said coding book, which have non-zero amplitude pulses, which respect the pre-established function. The system as described in Claim 19, further characterized in that the means of the pre-establishment function comprises means for pre-assigning, through means of the function Sp, one of the possible amplitudes q in the valid amplitude form for each position p, and wherein the pre-established function is respassed when the non-zero amplitude pulses of a combination of amplitude / position pulses each have an amplitude equal to the pre-assigned amplitude, by the function Sp to the position p of said nonzero amplitude pulse. The system as described in Claim 20, further characterized in that the means for pre-assigning one of the possible amplitudes q for each position p comprises: means for processing the speech signal to produce a signal D of the emission surface and a residual tone-removed signal R '; means for calculating an estimate of the amplitude of the vector B, in response to the signal D of the emission surface filtered backward, and to the residual signal R 'of tone-removed; and means for quantifying, for each of said positions p, an amplitude estimate Bp of said vector B, until obtaining the amplitude that will be selected for said position p. The system as described in Claim 19, further characterized in that said means for calculating an estimate of the amplitude of the vector B comprises means for adding the signal D of the emission surface in the normalized form: (l -ß) - to the residual signal R 'of tone-removed in the normalized form: R 'ß - 1 * 1 until obtaining in this way, an estimate of the amplitude of vector B of the form: B = (l> - £ + ß - Id | *; | where ß is a fixed constant. 23. The system as described in Claim 22, further characterized in that, ß is a fixed constant having a value between 0 and 1. 24. The system as described in Claim 23, wherein said means of quantization comprise means for quantifying, for each of said positions p, an estimate of the nominal peak amplitude Bp of said vector B, using the expression found below: Bp / max | Bn | where the denominator max | B n is a normalization factor, which represents a peak amplitude of the non-zero amplitude pulses. The system as described in Claim 19, further characterized in that said pulse combinations each comprise a number N of non-zero amplitude pulses, said apparatus additionally comprising means for restricting the p positions of the non-zero amplitude pulses, in accordance with at least one simple pulse permutation code plugged-N. 26. The system as described in Claim 22, further characterized in that said combinations of amplitude / position pulses each comprise a number N of non-zero amplitude pulses, and wherein the scanning means comprises means for maximizing a determined proportion having a denominator ak2, and means for computing said denominator a, through closed circuit means plugged-N according to the relationship that follows: + ü (P2 »Pl> + 2C7 '(p?, p2) + Í7' (p3, p3> + 2a / (p1, p,) + 2t7 '(p2, p3) ... •••. .... ^ '(pM, PH) * 2U' (pl, pt,) * 20f (p2tpH) + ... + 2U '(pH.l, PH) wherein the computerization for each closed circuit is written on a separate line from a closed outermost circuit to a closed innermost circuit of the closed circuits plugged-N, where pn is the position of the nonzero amplitude pulse ntn of the combination, and where U '(px, p and p) is a function dependent on the pre-assigned Spx amplitude up to a px position, between the p positions and the pre-assigned Spy amplitude up to the p position between the p positions. 27. The system as described in Claim 26, further characterized in that said means for calculating the denominator ak2 comprises means for jumping at least the innermost closed circuit of the closed circuits plugged-N, provided that the inequality which is below is true where Spn is the amplitude pre-assigned to the position pn, Dpn is the pntn component of the vector D of the emission surface, and Tp is a threshold related to the vector D of the emission surface filtered backward. EXTRACT OF THE INVENTION A coding book is explored according to the coding of a sound signal. This coding book consists of a group of combinations of amplitude / position pulses each defining different L positions and comprising both the zero amplitude pulses and the non-zero amplitude pulses assigned to the respective positions p = 1, 2 L. the combination, wherein each of the non-zero amplitude pulses assumes at least one of the possible q amplitudes. To reduce the complexity of the scan, a sub-group of amplitude / position pulse combinations of the encoding book is pre-selected in relation to the sound signal and only this sub-group of combinations is scanned. The pre-selection of the sub-group of combinations consists of the pre-establishment of a function Sp between the positions p = 1, 2, ..., L respectively, and the possible q amplitudes, in relation to the sound signal, being limited exploration to combinations of the coding book that have non-zero amplitude pulses, which respect the pre-established function. The function can be pre-established, by pre-assigning one of the possible amplitudes q to each position p, the pre-established function being respected when the nonzero amplitude pulses of a combination each have an amplitude equal to the amplitude Sp pre-assigned to the position p of said impulse.