FR2484755A1 - Multifrequency code signal receiver - has two non-recursive digital filters with pulse responses equal to cosine and sine functions multiplied by time window function - Google Patents

Abstract

The receiver comprises digital filters whose passband centres on the code frequencies, multiplying signal samples by filter coefficients. A circuit measures the power of the filter output signals. The digital filters are two non-recursive filters having step responses equal to a cosine and sine sample multiplied by a time window function respectively, the responses being in quadrature throughout the filter passband. The signal samples are in compressed code and the filter multiplication unit comprises a circuit forming address words composed of filter coefficient bits in uniform code and of at least certain compressed code sample bits. A memory is used to contain words each of which at given address is the product of the filter coefficient and the sample.

Description

MULTIFREQUENCY SIGNALING UNIFORM RECEIVER USING
NON-RECURSIVE FILTERS WITH WINDOW FUNCTIONS
The present invention relates to a multifrequency signaling receiver and, more particularly, to a non-recursive digital filter multifrequency signaling receiver.

The analog multifrequency signaling codes are internationally standardized. We recall below
the main features of these codes.

Keyboard multifrequency code is defined in the notice
Q 23 CCITT The signals are divided into two groups of four frequencies each, a code corresponding to the emission of two frequencies including one in each of the groups. Table I below gives the sixteen frequency pairs corresponding to the sixteen codes.

 The multifrequency codes S.O, C.O.T, E.L. and C.C.I.T.T.

R 1 are respectively composed of 2 frequencies out of 5, or of 2 of 6 frequencies. In the multi-frequency code S.O.C.Q.T., L, a sixth frequency, the control frequency of 1900 Hz, makes it possible to slave the exchanges between the applicant and the requested party. In the code R 1, two additional code signals have been introduced (KPet ST, start and end of the numbering) making a sixth frequency of 1700 Hz necessary.

Table II below gives the two frequency pairs forming the ten DTM frequency codes. and the twelve frequency pairs forming the sixteen codes R 1.

 The multifrequency code C.C.I.T.T. R 2 is applicable to two-wire circuits because of the separation into two groups of six frequencies, one for the forward and the other for the backward signals. The codes ensure, in addition to their own meaning, an enslavement of the exchanges between the applicant and requested. Each code is composed of two frequencies out of six. Table III below gives the combinations of the code R 2.

 Multifrequency receivers include a guard circuit that measures power in the voice band excluding signaling frequencies. If the noise detected by the guard circuit is excessive, the multifrequency receiver is inhibited to prevent any faulty operation.

The probability of simulation of a code number by noise is further reduced by precise timing of code numbers and intervals between code digits. Only signals satisfying the timing conditions produce valid definitive signals.

 As a result, depending on whether they meet the power and timing criteria or not, the signals are considered present or absent. More precisely, the parameters considered are the following - the frequency of the signals - the absolute and relative levels of the two tones constituting the code signal - the delay times at the recognition and the fallout of the receiver, the durations emission of the spurious signals or the possible cut-off times in a valid signal - the noises that can affect the white noise noise connection, parasitic lines (tones, distortion, intermodulation ...), imitation of codes by speech signals .

 The parameters defining the present signal state and the absent signal are summarized in Tables IV and V for the code R 2.

In Table V, q is the total number of signaling frequencies; q = 12 in the code R 2,
It is therefore necessary that the filters used make it possible to know the amplitude of the filtered signal or what amounts to the same power.

 Digital multifrequency signaling receivers using spectral analysis have been proposed in the prior art.

 In the time-division multiplex switching system type E 10 (see System E 10, the Equipment of Tones and Auxiliaries, by H. CAMPAGNO, J. JONCHERE, and J. TUAVDEN, "Switching and Electronics, NO 59, Oct. 1977, pp. 99-115), the multifrequency receivers comprise two second-order recursive digital bandpass filters, a detector and a first-order recursive digital low-pass filter, requiring a total of nine digital multipliers.

We also know (see "Digital MF Receiver Using Discrete Fourier Transform" by Ivan KOVAL and Georges GARA, IEEE
Transactions on Communications, VoL. COM-21, No. 12, December 1973, pages 1331 to 1335), multifrequency receivers comprising a discrete Fourier Transform computer of digital samples in which the Transform contains an integral core formed by a window time function.

The calculator of this reference calculates the real part and the imaginary part of a Fourier Transform Discrete1 the time function, constituted by samples, being multiplied by a window time function in the form of an integral core. The real part and the imaginary part are, of course, orthogonal for one or more frequencies but are not orthogonal for all the frequencies of a frequency band. On the other hand the proposed calculator is not a non-recursive filter.

<Tb>

<SEP> FREQUENCIES <SEP> LOWER <SEP> FREQUENOES <SEP> SUPERIOR
<tb> CODE <SEP> 697 <SEP> 770 <SEP> 852 <SEK> S41 <SEP> 1209 <SEK> 1336 <SEQ> 1477 <SEP> 1633
<tb><SEP> 1 <SEP> X <SEP> X
<tb><SEP> 2 <SEP> X <SEP> X
<tb><SEP> 3 <SEP> X <SEP> X
<tb><SEP> 4 <SEP> X <SEP> X
<tb><SEP> 5 <SEP> X <SEP> x <SEP> X
<tb><SEP> 6 <SEP> x <SEP> X
<tb><SEP> 7 <SEP> X <SEP> X
<tb><SEP> 8 <SEP> X <SEP> X
<tb><SEP> 9 <SEP> x <SEP> x <SEP>
<tb><SEP> 0 <SEP> X <SEP> X
<tb><SEP> * <SEP> X <SEP> X
<tb><SEP>#<SEP><SEP> X <SEP> X
<tb><SEP> A <SEP> X <SEP> X
<tb><SEP> B <SEP> X <SEP> X
<tb><SEP> C <SEP> X <SEP> X
<tb><SEP> D <SEP> X <SEP> X
<Tb>
TABLE I
KEYBOARD CODE

<tb><SEP> FREQUENCIES
<tb> CODE <SEP> 700 <SEP> 900 <SEP> 1100 <SEP> 1300 <SEP> 1500
<tb><SEP> 1 <SEP> X <SEP> X
<tb><SEP> 2 <SEP> X <SEP> X
<tb><SEP> 3 <SEP> X <SEP> X
<tb><SEP> 4 <SEP> X <SEP> X
<tb><SEP> 5 <SEP> X <SEP> X
<tb><SEP> 6 <SEP> X <SEP> X
<tb><SEP> 7 <SEP> X <SEP> X
<tb><SEP> 8 <SEP> X <SEP> X
<tb><SEP> 9 <SEP> X <SEP> X
<tb><SEP> 0 <SEP> X <SEP> X <SEP> 1700
<tb><SEP> KP * <SEP> X <SEP> X
<tb><SEP> ST * <SEP> X <SEP> X
<Tb>
* :: specific to the R1 system
BOARD #
MULTIFREOUENCE CODES SOCOTEL AND R1

<tb><SEP> SIGNALS <SEP> TO <SEP> THE BACK <SEP> SIGNALS <SEP> TO <SEP> THE FRONT
<tb> CODE <SEP> 540 <SE> 660 <SE> 780 <SE> 900 <SE> 1020 <SE> 1140 <SE> 1380 <SE> 1500 <SE> 1620 <SE> 1740 <SE> 1860 <SEP > 1980
<tb><SEP> 1 <SEP> X <SEP> X <SEP> X <SEP> X
<tb><SEP> 2 <SEP> X <SEP> X <SEP> X <SEP> X
<tb><SEP> 3 <SEP> X <SEP> X <SEP> X <SEP> X
<tb><SEP> 4 <SEP> X <SEP> X <SEP> X <SEP> X
<tb><SEP> 5 <SEP> X <SEP> X <SEP> X <SEP> X
<tb><SEP> 6 <SEP> X <SEP> X <SEP> X <SEP> X
<tb><SEP> 7 <SEP> X <SEP> X <SEP> X <SEP> X
<tb><SEP> 8 <SEP> X <SEP> X <SEP> X <SEP> X
<tb><SEP> 9 <SEP> X <SEP> X <SEP> X <SEP> X
<tb> 10 <SEP> X <SEP> X <SEP> X <SEP> X
<tb><SEP> 11 <SEP> X <SEP> X <SEP> X <SEP> X
<tb><SEP> 12 <SEP> X <SEP> X <SEP> X <SEP> X
<tb><SEP> 13 <SEP> X <SEP> X <SEP> X <SEP> X
<tb><SEP> 14 <SEP> X <SEP> X <SEP> X <SEP> X
<tb><SEP> 15 <SEP> X <SEP> X <SEP> X <SEP> X
<Tb>
TABLE III
CODE R2

<tb><SEP> R2
<tb><SEP> FREQUENCY <SEP>#<SEP> 10Hz
<tb><SEP> Mini, <SEP> Max, <SEP> 35dBm <SEP> -4dBm
<tb> LEVEL
<tb><SEP> Difference <SEP> between <SEP> two <SEP> lines <SEP> 7dB
<tb><SEP> Minimum <SEP> minimum
<tb><SEP> Time <SEP> minimum <SEP> between
<tb><SEP> two <SEP> codes
<tb><SEP> delay <SEP> cumulative <SEP> to <SEP> la
<tb> TIME <SEP> detection <SEP> and <SEP> a <SEP><SEP> 80
<tb><SEP> fallout
<tb><SEP> Breaks <SEP> max.
<Tb>

<SEP> ~ Difference <SEP> between <SEP> the appearance
<tb><SEP> of <SEP> two <SEP> frequencies
<tb><SEP> Noise <SEP> white <SEP> ~ 42 <SEP> dBm
<tb><SEP> Stripe <SEP> (s) <SEP> parasites
<tb><SEP> NOISES
<tb><SEP> Intermodulation <SEP> 1 <SEP> to <SEP> 9-2 <SEP> Frequency <SEP> of <SEP> Code <SEP> to <SEP> ~ 20dB
<tb><SEP> (global) <SEP> / <SEP> Freq. <SEP> of <SEP> plus <SEP> Strong <SEP> level
<tb><SEP> Distortion
<Tb>
TABLE IV
STATE SIGNAL PRESENT

(1) expression dans lesquelles
0 # n # N - 1 i i est la fréquence angulaire à reconnattre #i = - (N - 1) #iT/2
T est la période d'échantillonnage (T = 125 fus pour un échantillonnage à 8 kHz) w(n)est une fonction de fenêtre définie par N échantillons.<tb><SEP> R2
<tb> WATER <SEP> Maxi <SEP> ~ 42 <SEP> dBm
<tb><SEP> LEVEL <SEP> Difference <SEP> between <SEP> two <SEP> lines <SEP> -20 <SEP> dB
<tb><SEP> TIME <SEP> Time <SEP> maximum <SEP> 7ms
<tb><SEP> Stripe <SEP> (s) <SEP> parasites <SEP> 2 <SEP> Freq <SEP> - <SEP> 5 <SEP> dB <SEP> m
<tb><SEP> [1300.3400Hz] <SEP> or <SEP> [330, <SEP> 1150]
<tb> NOISE
<tb><SEP> and <SEP> [2130,3400]
<tb><SEP> next <SEP> receiver <SEP>
<tb> TABLE V
STATE SIGNAL ABSENT
The multifrequency signaling receiver of the invention comprises pairs of narrow bandpass filters for each signaling frequency. These filters, designated respectively as a cosine filter and a sine filter, are non-recursive filters of order N defined by their impulse responses finished as follows
Ci (n) = w (n) cos (n #iT + #i)
If (n) = w (n) sin (n #iT + #i)

(1) expression in which
0 # n # N - 1 ii is the angular frequency to be recognized #i = - (N - 1) # iT / 2
T is the sampling period (T = 125 fu for 8 kHz sampling) w (n) is a window function defined by N samples.

Because of the value of. and the symmetry of the window functions, the sequences C (n) and Si (n) are respectively
They are symmetrical and antisymmetric, which means that both filters are quadrature across the spectrum (their frequency responses have a constant phase shift of # / 2).

In addition, it is possible to demonstrate that the frequency response modules of the 2 filters have quasi-identical bandwidth characteristics, directly related to the shape of the Fourier Transform of the window (central lobe width, lobe level secondary).

The previously mentioned properties make it possible to estimate at the 1 / NT rate the power of frequency oscillation #ii and of unknown phase by processing N samples of the input signal as follows
1) calculate the Nine output samples of the two non-recursive filters by the convolution products

où x(k) représente le kième échantillon du signal d'entrée
2) obtenir ensuite la puissance P.

where x (k) represents the kth sample of the input signal
2) then obtain the power P.

2Pi = yci (N-1) 2 + ysi (N-1) 2 (2 ')
Note that due to the symmetry and antisymmetry of the C (n) and S (n) responses, equations (2) can be
ii replaced by the expression (2 ') without modifying the value of Pi

The type of window function chosen and the number N of samples processed for each estimate completely defines the frequency responses of the filters; the larger the N, the better the frequency selectivity of the filters for the same window function. It should be noted, however, that in fixing the duration of the elementary spectral analyzes the value of N must be chosen in accordance with the time specifications such as maximum recognition time (slave codes), minimum lengths of digit and inter-digit (code keyboard)
Non-recursive filters are filters in which all multiplications relate to the single input signal.

No multiplication is made on a signal which would be deduced from the input signal by an addition. As a result, according to another characteristic of the invention, the non-recursive filtering is performed on the compiled samples taken from the digital multifrequency signal without the need to decompress the samples before multiplication.

 the single multiplier in each non-recursive finite impulse response filter in cosine and sinus with integral window core comprises concatenation circuit this concatenation circuit providing from the codes representative of a filter coefficient and a compressed sample address of a word to read in the memory. This word is the product of said coefficient by the uncompressed sample.

 In order to make this feature of the invention clear, the notions of uniform code and compressed code will be recalled.

Uniform quantization (gu) and uniform code (cod u) assume that there are M = 2 uniform quantization levels which or 0 # @ # M - 1
These uniform quantization levels all-lower than unity are that at M-1 and the level that is equal to (# + 1/2) / M
At each quantization level is associated a code, the quantization level where @ = (# + 1/2) / M is coded by the integer #
If, for example, = 6, there are sixty four levels of quantification
0 to 63
what are respectively worth
1, 3, 5, .. 93. ,, 125, 127 128 128 128 128 128 128 and are coded respectively by the codes (in binary numbering)
O, 1, 2, ..., 46, .., 62, 63
A sample x strictly less than 1 is coded uniformly by
cod u (x) = Integer (xx M) and its quantization value is
u (x)
that
If for example x = 0.732 and p = 6 (7-bit coding, including sign), we have cod u6 (0.732) = Integer (0.732 x 26) = 46 and the quantization value is

Compressed quantization (qc) and compressed code (cod c)
We define the compression law by
y = f (x) and we still suppose that there is
M = 2 levels of compressed quantization qc # where # has the same boundaries as before.

These compressed quantization levels all lower than unity are and the level

<tb><SEP> O <SEP> Ml
<tb><SEP> qc <SEP> to <SEP> qc
<tb> p
<tb> qc <SEP> is worth
<tb><SEP> p <SEP> f-1
<tb><SEP> qcp <SEP> = <SEP> 0 (6 <SEP> + <SEP> 1/2) / Mg
<Tb>
At each quantization level is associated a code, the quantization level
f-1 [(# + 1/2) / M] being coded by the integer 5
A sample x strictly less than 1 is encoded with compression by
cod cp (x) = Integer Lf (x) x M] and its quantization value
cod cp (x)
qc
If we take as an example the law A recommended by the
CCITT and defined by thirteen straight segments of which 7 for the positive samples and 7 for the negative samples, one of the segments being common to the samples of the two polarities, we have

<tb><SEP> 16 <SEP> x <SEP> for <SEP> x <SEP><1/64<SEP>
<tb><SEP> 8x <SEP> + <SEP> 1/8 <SEP> 1/64 <SEP><SEP> x <SEP><SEP> 1/32
<tb><SEP> 4x <SEP> + <SEP> 1/4 <SEP> 1/32 <SEP><SEP> x <SEP><SEP> 1/16 <SEP>
<tb><SEP> f (x) <SEP> = <SEP>#<SEP><SEP> 2x <SEP> + <SEP> 3/8 <SEP> 1/16 <SEP><<SEP> x <SEP><<SEP> 1/8
<tb><SEP> x <SEP> + <SEP> 1/2 <SEP> 1/8 <SEP><SEP> x <SEP><SEP> 1/4
<tb><SEP> x / 2 <SEP> + <SEP> 5/8 <SEP> 1/4 <SEP><SEP> x <SEP> 4 <SEP><SEP> 1/2
<tb><SEP># x / 4 <SEP><SEP> + <SEP> 3/4 <SEP> 1/2 <SEP> - <SEP> x <SEP> I <SEP>
<tb><SEP> y / 16 <SEP> y <SEP> 41/4 <SEP>
<tb><SEP> y / 8 <SEP> - <SEP> 1/64 <SEP> 1/4 <SEP><<SEP> y <SEP><<SEP> 3/8
<tb><SEP> y / 4 <SEP> - <SEP> 1/16 <SEP> 3/8 <SEP><<SEP> y <SEP><<SEP> 1/2
<tb> f-1 (y) = <SEP># y / 2 <SEP><SEP> - <SEP> 3/16 <SEP> 1/2 <SEP><<SEP> y <SEP><<SEP> 5/8
<tb><SEP> y <SEP> - <SEP> 1/2 <SEP> 5/8 <SEP><<SEP> y <SEP><<SEP> 3/4
<tb><SEP> 2y <SEP> - <SEP> 5/4 <SEP> 3/4 <SEP>(<SEP> y <SEP><<SEP> 7/8
<tb><SEP> 4y <SEP> - <SEP> 3 <SEP> 7/8 <SEP><SEP> y <SEP><SEP> 1
<Tb>
S x = 0.135 and = 7 (encoding in 8 bits including sign) we have
cod c7 (0.135) = Integer [(0.135 + 1/2) x 27 # = 81 and its quantization value is
qc81 = f-1 [81.5 / 128] = 0.13671875
7
Application to digital filtering
Assuming that the filter coefficients are uniformly coded on 6 bits (7 bits including sign) and the 7-bit coded signal samples (8 bits including sign) with CCITT law A compression a memory of 8K words (6 + 7 = 13 bits of addressing) is necessary for.

achieve the multiplication of a filtering coefficient,
T by a signal sample, ie o
For example suppose T = 0.732 and 0.135
T is the exact value of a coefficient, deduced for example from the calculation of expressions (1)
a is the value of the telephone channel sample before
coding
In this case we find - in the memory used to store the filtering coefficients on 6 bits (7 with the sign) the code code u6 (T) = T = 46 - on the telephone channel to be processed, scanned on 7 bits (8 with the sign) the cod code c7 (a) = Z = 81
The address of the multiplication memory formed by concatenation of the bits representing T and # is in this case
# + 27 T = 81 + 128 x 46 = 5969
In other words, # represent the seven least significant bits of the address and T the six bits of weight .fort. from this address.

The word of address
5969 = # + 27 T must be representative of the product
## = 0, 135 x 0.732
= 0.09882
Due to the loss of information on the exact values of a and T resulting from the quantization operations this word will represent the product of the respective quantization values cod c7 (a) cod u6) T)
qc7 x qu6
= 0.1391875 x 0.7265265
= 0.099334717
If the words of the multiplication memory consist of 8 bits, one will find at address 5969 the number code u8 (0.099334717) = 25 representative of the quantization value qu82 = 0.099609375
It is possible, at the cost of an acceptable increase of the quantization noise, to amputate the binary number representing the telephone sample of a number of low-order bits.
By way of example, the case where the compressed code is normalized on 7 bits (8 bits including sign) is reduced by 2 bits of low weight: the samples are then coded on 5 bits (6 bits including sign) and it is important to determine the new quantization levels to associate with the amputated compressed codes.

In the case previously treated, a = 0.135 was associated on 7 bits (8 bits including sign) the code
z = cod c7 (#) = 81 and the quantization level
81 7 qc7 = 0.1367755. With 5 bits the amputated code # 'is 200 The method that minimizes the quantization error is to associate with the amputated code #' the quantization level = f-1 [20.5 / 32] = 0.140625
If we always keep 6 bits (7 bits including sign) to code the coefficients, the amputation of 2 bits of the compressed codes makes it possible to reduce to 2K words (6 + 5 = 11 bits of addressing) the size of the memory multiplication. The way of filling this memory can be illustrated as previously with the following example coefficient T = 0.732 uniform code on 6 bits T = 46
level of quantification
qu646 = 0.7265625 signal a = 0.135 7-bit compressed code Z = 81
amputated code of 2 bits (5 bits)
# '= 20
level of quantification
quc520 = 0.140625 memory address: # '+ 25 x T = 20 + (32 x 46) = 1472 content (on 8 bits) code u8 (qu646 x qc52Q) = 26 representative of the quantization value qu86 = 0.103515
The invention will now be described in detail in connection with the accompanying drawings in which - FIG. 1 depicts in block diagram form the two non-recursive filters forming the multifrequency signaling receiver; FIG. 2 represents the frequency response of the filters using a window function; and - FIG. 3 represents in the form of a block diagram, the filter multiplication circuit.

 Referring to FIG. 1, the samples of the multifrequency signaling signal are applied to the common input 10 of two non-recursive digital filters 1 and 1.

cs
Each of these filters has an input multiplier 11
c and 11, an adder 12 and 12 and a delay circuit
scs 13 and 13 located between the exit and the entrance of each addi
cs
The outputs of the adders 12 and 12 are connected
cs to a squaring circuit, respectively 14 and 14
cs through an AND gate, respectively 15 and 15
cs
The outputs of the elevation circuits squared 14 and 14 are
cs connected to an addition circuit 16 itself connected to a multifrequency signal recognition logic circuit 17.

The latter receives the outputs of other pairs of recursive filters corresponding to the other frequencies of the multi-frequency code considered.

A time base 18 controls AND gates 15 and 15
cs 19, 19 and 19e. She opens the first two at the pace
cse
NT and the last three at the T pace
The memory 20 comprises two parts 20 and 20, one
cs, containing the samples of ai (n) and the other containing the samples of S. (n). The sampled functions C (n) and S (n) were plotted next to the memories 20 and 20.
csii in the case where w (n) is a Kaiser-Bessel window defined in the article "On the Use of Windows for Harmonic Analysis with the Discrete Fourier Transform" by Fredric J. HARRIS,
Proceeding of the IEEE, Vol. 66, No. 1, January 1978, pp. 73-74.

 Fig. 2 represents the frequency response of the sine filter for 1300 Hz.

 The quantification of the coefficients of the filters does not modify the phase properties of the frequency responses because it retains the symmetry and the antisymmetry of the impulse responses. Only the modules are assigned A format of seven-bit words is satisfactory because it only alters the sidelobes below 40 dB tPiy, 2).

 Referring to FIG. 3 shows only the multiplier and the adder of the cosine impulse response filter. The filter coefficients are uniformly coded on 7 bits, ie 6 absolute value bits and 1 sign bit, and the samples are coded with 8 bit compression, ie 7 absolute value bits and 1 sign bit.

 The multiplier of the filter essentially comprises a concatenation register 21 and a so-called multiplication memory 22.

The 7-bit filtering coefficients stQ.cked in the coefficient memory 20c are transferred under the control of the time base 18 and the AND gates 19 ~ into the positions
c
O to 5 and 6 of the concatenation register 21 Samples
s
adu 9-bit multifrequency signal received on the telephone channel are transferred under the control of the time base 18 and AND gates 19é, in positions 7 to 13 and 14s of the concatenation register 21.

The positions 6 and 14 of the register 21 are connected to
ss an exclusive OR gate 23 which produces the product of the signs of the coefficient and the sample. Positions 7 and 8 are unbound and the bits that go into them are not used
This lack of connection is the amputation of two bits discussed above. The positions 0 ak 5 inclusive and 9 to 13 inclusive are connected to the addressing device of the multiplication memory 22. They convey the addresses to Il bits
# + 25 x T that was discussed in the introduction
The 8-bit words read at the read output 24 are applied to the adder 25. The latter is part of a loop having a cycle duration equal to the sampling period and which comprises the accumulator 26.

 The time base 18 controls the filling of the register 21, the operation of the accumulator 26 at the rate T and the opening of the door 15 at the rate NT.

c
The sign bit of the sign product is applied to the adder 25.

The number of coefficient bits, samples, addresses and filtered products are given on the
Fig. 3 as an example and can be modified without departing from the scope of the invention. But it must be emphasized that
the samples applied to the concatenation register and the multiplication memory as an address formation factor are compressed samples

Claims (4)

 1 - Receiver of digital multifrequency code signaling signals comprising digital bandpass filters centered on the code frequencies, multiplying samples of the signaling signal by filtering coefficients and means for determining the power of the output signals of said filters, characterized in that the digital filters centered on a given code frequency are two non-recursive filters (1c, 1s) respectively having impulse responses equal to a sampled cosine and sine multiplied by a window time function, said responses being in quadrature throughout the bandwidth of the filters.  2 - digital multifrequency code signaling signal receiver according to claim 1 wherein the samples of the signaling signal are compressed coded code samples, characterized in that the filter multiplication unit which multiplies the samples in code compressed by the uniform code filter coefficients comprises a concatenation circuit forming address words composed of the bits of the uniform code filtering coefficients and at least some bits of the compressed code samples, and a memory containing words of memory, the memory word at a given address being the filter coefficient product by the sample.  3 - Receiver of multifrequency code digital signaling signals according to claim 2, characterized in zen that the samples in compressed code have 8 bits, one of which has a sign and 7 of absolute value and the uniform code filter coefficients have 7 bits including one sign and 6 absolute value and the address words formed by the concatenation circuit have 13 bits.  4 - Receiver of multifrequency code digital signaling signals according to claim 2, characterized in that the compressed code samples have 6 bits, one of which is of sign and of absolute value and the uniform code filtering coefficients have 7 bits of which one of sign and 6 of absolute value and the address words formed by the concatenation circuit have 11 bits.