FIG. 1, (a,b and c) illustrates a prior art frequency band saving method.

FIG. 2 (a,b, and c) illustrates a frequency band saving method of the present invention for comparison with FIG. 1.

FIGS. 3,4 and 5 illustrate, respectively, the waveforms in (1) a system where both high and low frequency waves are transmitted, (2) a system as illustrated in FIG. 1 where slices or segments of both high and low frequency waves are transmitted, and (3) a system as in FIG. 2 where the low frequency component is totally transmitted and the high frequency component is transmitted in segments or slices.

FIG. 6 is a schematic block diagram of an apparatus for carrying out the process illustrated in FIG. 5.

FIG. 7 (a to f) is a view similar to FIG. 2 but showing a modified system of transmitting the signals in a to c, and the system of restoring the received signal in d-f.

FIG. 8 is a schematic block diagram of apparatus for carrying out the system of FIG. 7.

FIG. 9 is a detailed schematic block diagram of the comonent ZE₁ of FIG. 8.

FIG. 10 is a detailed schematic block diagram of the component ZK₂ of FIG. 8.

FIG. 11 is a view similar to FIG. 2, but illustrating another modified method or system of the invention.

FIG. 12 is a schematic block diagram of an apparatus for carrying out the process illustrated in FIG. 11.

FIG. 13 (a-d) illustrates another modified system of the invention.

FIG. 14 is a schematic block diagram of an apparatus for carrying out the process illustrated in FIG. 13.

FIG. 15 (a--c) illustrates still another modified system of the invention.

FIG. 16 is a schematic block diagram of an apparatus for carrying out the method of FIG. 15.

A known process of this kind is illustrated schematically in FIG. 1; in FIG. 1a one sees the original segments A₁, A₂, . . . with the length, T_o, FIG. 1b shows the reduced segments A_1r, A_2r . . . and in FIG. 1c one sees the received signal with the gaps L of FIG. 1b filled by the repetition of reduced signals A_1r ', A_2r '. . . .

This process has the drawback that, because of the regularly spaced discontinuities in the reconstructed signal, there exists disagreeable amplitude and phase shifts with the following effects:

additional disturbing noise

diminished understanding

more difficult understanding of the basic frequency (pitch frequency) and thereby unnatural sound and loss of speech recognition.

The invention is based on the problem of eliminating these drawbacks and of preventing unnecessary redundancy so that the transmission will be possible with reduced bandwidth or reduced time without noteworthy loss in quality (in regard to understandability and audible noise). The invention is based on the unexpected discovery that the suppression of part of the segments and their replacement by preceeding segments operates especially disadvantageously on the transmission quality at the low frequencies, while the effects at the higher frequencies are insignificant. According to the invention, the above-mentioned drawbacks of the known process, therefore, are largely prevented by splitting the signal to be transmitted into at least two frequency bands, the lowest frequency band being transmitted in its entirety, while at least one other frequency band of a higher frequency structure is split up through periodic interruption into short signal segments or slices and wherein at the receiving end, the gaps of these short segments are filled by repetition of the transmitted segments or slices.

The new process will be compared with the old one and further explained with the aid of FIGS. 1-5.

According to the known process, the signal with the bandwidth F_a (see FIG. 1a) is split up into the segments or slices A₁, A₂ . . . with the period T_o. In the further figures, the capital letters (A₁,A₂ . . .) will show the designated blocks as signal segments, wherein the height of the block corresponds to the range of the frequency. For designating the instantaneous value of the periodic changing signal, in contrast, small characters are used. The segments will, according to FIG. 1, be shortened through interruptions of segments L, so that only the segments A_1r,A_2r . . . are transmitted. At the receiving end, the gaps as shown in FIG. 1c are filled by repeating the transmitted segments (designated A_1r ', A_2r ' . . .) so that an uninterrupted signal results. In the gaps additional signals may be transmitted.

According to the invention, the signal as shown in FIG. 2a will first be split into at least two bands with the segments A₁,A₂ . . . and B₁,B₂ . . . whose bandwidth is indicated by F_a and F_b. The lower frequency band is transmitted unchanged, as shown in FIG. 2b. With the upper frequency band, the segments, on the contrary, are interrupted partially so that only the reduced partial segments B_1r, B_2r . . . are transmitted. On the receiving side, the gaps in the upper band, as shown in FIG. 2c, are filled by repeating the transmitted segments (B_1r ',B_2r. . .). The segment length T_o is so measured that the essential characteristics of the speech signals at any given time are longer than the interruption so that they are not completely lost through the blanking. Thus, the segment length is about 20-50 milliseconds.

To further illustrate, transmission of two sine waves a and b is shown in FIG. 3, wherein there is a lower frequency band A and an upper frequency band B. These signals are divided into the signals a₁,a₂ . . . or b₁,b₂ . . . with the duration T_o which are longer than the period T_a or T_b of the sine waves. According to the known process explained in connection with FIG. 1, these sine waves would both be periodically interrupted so that, as shown in FIG. 4, only the reduced segments with the duration T_0/2 are transmitted, which results in the signals a_1r and b_1r as well as a_2r and b_2r, etc. Through the receiving side repetition, the full signal is formed with a_1r ' and b_1r ' as well as a a_2r ' and b_2r ' etc., filling the gaps. The low frequencies, as well as the high frequencies are not continuous at the points of sudden change S_a or S_b because the complete signal is generally not synchronized with the delayed signal. The noise appearing through such sudden changes increases as, the number of undisturbed periods lying therebetween decreases. As a measure of the noise therefore, the quantity q can serve where

q = 1/z = nTx/To 

where Z is the period number of the wave per reduced segment, n is the proportion between the segment length to the reduced segment and T_x is the period length of the signal. With a segment length T_o = 20 ms. and a reduced segment length, T_o /n, of 10 ms. the proportion n = 2 and it follows that for sinusoidal tones of 300 Hz and 3000 Hz, the following period lengths T_x and noise density q_x exist:

t₁ = 300 Hz;

T₁ = 1000/300 = 3.3 ms.;

q₁ = 6.6/20 = 0.33.

t₂ = 3000 Hz;

T₂ = 1000/30000 = 0.33 ms;

q₂ = 0.66/20 = 0.033.

Above all, it is clear that the detrimental irregularities at the low frequencies should be diminished. This may be attained according to the invention by transmitting the lower frequencies directly without interruption and without supplementary filling in any gaps as shown in FIG. 5; the low frequency signal, a, therefore shows no irregularities. The signal b, which is in a higher band of frequencies retains, on the contrary, the irregularities S_b which, as explained above, produce little noise. Besides, it should be noted that in the higher range of frequencies, the non-sinusoidal forming occurrences (hisses and explosive noises) is considerably greater and that moreover, from experience, the ear is considerably less sensitive in the range of higher frequencies with respect to small deviations. Through direct transmission of low frequencies, it will also result that those frequencies (especially the fundamental frequency or pitch frequency), which are especially important for recognition of a voice, permit satisfactory comprehension thereof, which is indispensible, not only for good intelligibility but also for speaker recognition. The gain produced through this process, that is, the expansion of the necessary transmission capacity, corresponds to the screened out segments L of FIG. 2, whose area is expressed by (T_o - T_b) F_b = (1 - 1/n) T_o F_b. The relative gain, p corresponds to the proportion based on the total area. ##EQU1## where several band groups are thinned out, and the blanking proportions n_b, n_c of the band groups are not equal then ##EQU2##

The width of the unthinned fundamental band A can be, for example F_a = 800 H_z (300-1100 H_z) and the width of the thinned out band can be F_b = 2400 H_z (1100-3500 H_z) whereupon, working with a blanking proportion of 2a = 3, the relative gain is then ##EQU3##

Through increased thinning of the upper frequency band or division into several bands on which the upper frequency bands come in for still greater thinning out, the gain can naturally be increased.

FIG. 6 shows an apparatus for carrying out the FIG. 2 type of process. The speaking signal s is divided by the band splitting network FW, into lower and upper frequency subbands, designated by signals a and b. The interrupter U takes care of suppressing the gap portions L (FIG. 2) so that only the remainder segments B_1r, B_2r . . . , as the thinned out signal for the upper band, arrive for transmission while the total signal a of the lower subband is transmitted. The transmitted signal g is distributed again at the receiving side by the band splitter FW₂ into two subbands of signals a and b_r. By a delay of the periodically interrupted signals, b_r, by the shift register SR or by a similar delay apparatus, the filling signal b_r is produced which fills the gaps of b_r. There results though such repetition, the part signal b_rw which comprises the upper frequency band F_b (FIG. 2) which together with the unchanged signal band a provides the final received exit signal s_w which has only slightly audible noise as is well understood.

The gaps L (FIG. 2), that is, the interrupted segments of the switch U (FIG. 6) can be used for transmitting other signals.

According to FIG. 7, another practical signal alteration for utilization of these gaps for decreasing the need of channel space is made possible. In FIGS. 7a and 7b, again the lower part of the frequency band and the individual segments are shown (as in FIG. 2) and the upper subband is divided into segments B_1r and B_2r . . . . The latter are reshaped through time expansion into segments shown in FIG. 7c which, at any given time, show a doubled length of halved frequency so that the gaps vanish. Since now, the lower margin of the upper bands have a frequency of F_a /2, the whole upper frequency band must, according to FIG. 7d, be lifted about an amount of F_a /2 in order to make possible a transmission of the unchanged lower band. The total band width amounts now to F_a + F_b /2 so that a channel with reduced width suffices for transmission of the resulting signals g. On the receiving side, the upper frequency subband is first reshifted into the low frequency subband, according to FIG. 7e (corresponding to FIG. 7c) and then the time compression of the upper frequency segments is effected so as to again reproduce the interrupted signals B_1r, B_2r . . . as in FIG. 7b. The repeating of these upper range frequency sections B_1r,B_2r . . . finally supplies the gap-free signal with the filler sections B_1r ',B_2r ' . . . as in FIG. 7f.

For carrying out this frequency band reduction, an apparatus, according to FIG. 8 serves wherein first again the signals s are divided into the subband signals a and b by means of the band splitter FW₁. A time expansion of the upper band signal in ZE, provides the signal b_c which comprises the band segments B_1e, B_2e . . . included in FIG. 7c, a frequency transformation in frequency transformer FU, again yields the signal b_eu corresponding to the sections B_1eu, B_2eu . . . in FIG. 7d. Through addition of the subband signals a, the total signal to be transmitted g results. The subband signal a is additionally delayed appropriately in VE, since the time expansion of the upper frequency band in ZE is accompanied by a delay. In this case, the partial band signals b_eu and a_v are brought together in better correspondence. After reception, separation of the signals to subbands b_eu and a_v by the band splitter FW₂, and then a reconversion of b_eu to the original frequency takes place in the frequency transformer FU₂, whereupon the time compression of the upper frequency subband b_e takes place in ZK₂ to again give signals b_rw. In connection with the time compression, these signals become suitable for the purpose of filling the gaps and one obtains thus the subband signal b_rw which comprise the band segments B_1r , B_1r ', B_2r, B_2r ' . . . of FIG. 7f. Since the compression is also associated with a time delay, a further delay of the lower frequency band is provided in VE₂ so that the lower frequency signal is shown as a_vv whose time lag again corresponds to that of the upper signal, whereupon the two signals together again finally give an understandable signal s_w with the original band frequencies.

An apparatus ZE for carrying out the time expansion (ZE of FIG. 8) is shown in FIG. 9. A subband segment of signals b is first conducted over the switch U₁ to the shift register SR₂ whose memory, through the block pulse e, has already stored the previous segment which now is taken out over the widthdrawal switch U₂. The withdrawal, as controlled through clock pulse e₂ (conducted over V₂) whose repetition frequency is slower than that of e, so that the withdrawal is correspondingly slower and results in correspondingly longer duration: the expanded band segment having lower band width corresponding to the longer duration.

With each band segment, the switches U₁,U₂ and also V₁,V₂ shift so that the charging and discharging of the shift registers repeats alternately whereby an upper band signal be with reduced frequency and reduced band width results.

For recovering the upper subband segments of the original band width, the time compressor, ZK₂ (FIG. 8), serves, which is carried out by the apparatus ZK of FIG. 10. The signal b_e in the form of a reduced subband segment B_r is conducted over switch U₄ to the shift register SR₄. Simultaneously, the shift register SR₃, which has already been loaded with the previous segment is discharged through switch U₅. The fast and the slow clock pulse sequences e₁ and e₂ are conducted to the register over the switches V₁,V₂ so that the discharging of the register is always faster than the charging. With the first discharge of registers SR₄, the outgoing signal is led back over switch U₄ to the entrance so the gaps produced by the past discharge are filled by the repeated signal. Also, with the compression apparatus, the function of SR₃ and SR₄ are exchanged through the reversal of switches U₃ and U₄ after each complete segment (i.e. of time space T_o) so that finally an uninterrupted signal b_rw results, which comprises the band segments B_1r, B_1r ', B_2r, B_2r ' corresponding to FIG. 7f.

The described time expansion and compression can also be carried out with other values of n wherein the proportion of the beat frequencies of e₁ and e₂ are related to n. A still more effective surplus reduction is obtained through subdividing the signal frequency band into 3 subbands, of which the two upper ones are thinned out through screening of part of the segments. It is advisable to foreshorten the upper band a greater amount (e.g. n₂ = 4) than the intermediate band (e.g. n₂ = 2) as shown in FIG. 11a. Then the multiple reproduction of the higher frequency signals is without significant influence on the quality of the transmission. The manner in which reduced segments B_1r, B_2r . . . produced by screening out portions of the band are transformed into a gap-free sequence of segments B_1eu, B_2eu of diminished band width by expansion and frequency transformation has already been explained. In similar ways the C_1r, C_2r . . . segments can be transformed to a gap-free sequence of segments C_1eu, C_2eu . . . . For this, naturally, the expansion proportion is adapted to the greater proportion of n₂. With this, the subband segments result in the signals to be transmitted as shown in FIG. 11b. A receiver time compression and repetition of the subband sections yields finally the uninterrupted sequence of subband segments in three frequency ranges as shown in FIG. 11c. Since the phase and amplitude transitions at the disconnected slices occur outstandingly in the upper subbands where they are slightly perceptible in the transmission quality, the enlarged gain is obtained without substantial loss of quality.

An apparatus for carrying out this process with subbands is shown in FIG. 12. The processing of the subband signals a and b corresponds to that of FIG. 8 and is only an extension thereof with the band splitters FW₃, FW₄ . . . for extraction of the subbands c of c_eu with the expansion and compression circuits ZE₃, ZK₄ and the frequency translators FU₃, FU₄ provided for processing these signals.

A process for the use of the subband gaps for transmitting an additional signal without time expansion and compression is shown in FIG. 13. The first signal is, according to FIG. 13a, transmitted according to the subband thinning shown in FIG. 2b. The second signal with the foreshortened sections B_1r *, B_2r *, . . . of the upper subband and unshortened segments A₁ *, A₂ *, . . . of the lower subband is transferred temporarily opposite to the first signal according to FIG. 13b around a half-section length T_o /2, and it will, by frequency inversion, be brought to the reversed position as in FIG. 13c. The two signals thus nested into one another as in FIG. 13d can be transmitted with only slightly enlarged total bandwidth.

The apparatus for carrying out this two channel transmission is shown in FIG. 14. The first signal s, corresponding to that of FIG. 6, is transmitted through band splitter FW₅, the upper subband being interrupted periodically at U₆ and, in the receiving end, band splitter FW₇ again splits out the subband signals a and b_r. A delay in SR₅ serves again for gap-filling so that finally a gap-free signal s_w, corresponding to the original bandwidth, results. The second signal s* undergoes first, in the inverter FI₁, a frequency inversion so that in the inverted signal s*, the lower subband appears as the upper subband, while the inverted upper subband is situated adjacent the upper subband of the first signal. A periodic switch-over of the subband signal b,b₁ * (which corresponds to the original upper band) leads to the resulting signals g for transmission of the structure recognizable in FIG. 13d. The original lower subband signal a₁ * of the second readjusted signal will be split out at the receiving end by a band splitter FW₇ and similarly the original upper subband signal b_ri *, after passing through switch U appears as in FIG. 13c with interruptions. The interruptions are filled in with the delayed segments by SR₆ so that an uninterrupted subband signal b_riw * results. This, combined with a*, passes to the frequency inverter FI₂ where it is finally brought to the original frequency range whereby an understandable exit signal s_w * results.

Another method for two channel transmission, according to a multiplex system, is shown in FIG. 15. A first signal according to FIG. 15a is split up into the segments A₁, A₂ . . . of lower subband and the upper subband segments B_1r,B_2r . . . foreshorted to T_o /4. Instead of time expansion of the foreshortened segments, a time compression of the unshortened lower subchannel segments A₁,A₂ . . . is provided. By this compression, a foreshortening to T_o /4 results which simultaneously increases all frequencies thereof by a factor of 4 so that the segments A_1k,A_2k . . . are in a similar frequency range and length as B_1r,B_2r as shown in FIG. 15b. The gaps between these segments suffice directly for transmitting another signal in the same way. At the receiving end as shown in FIG. 15c, the lower subband is received through time expansion of the segments A_1k,A_2k . . . , while the reduced segments of the upper band are obtained through multiple repetition to give an uninterrupted sequence of segments B_1r ¹,B_1r ² . . . B_2r ¹,B_2r ².

The carrying out of this two channel transmission can ensue from the very simple arrangement shown in FIG. 16. The subband signals a and b of transmission signal s obtained from the band splitter FW₈ are compressed on the one side by time compressor ZK₈ to provide shortened signal segment a_k and on the other side the signals b pass through interrupter U₈ to produce the foreshortened segments b_r, whereby the length and breadth of a_k and b_r synchronize. In like way, a second transmission signal is processed that finally fills in the gaps of g. On the receiving side, a splitting of the signal segments a_k,b_r which follow one another in sequence is accomplished by switch U₉. The lower frequency subband a is obtained through time expansion of signals a_k by ZE₉. The uninterrupted sequence of the upper subband is obtained by repetition of the signal segments b_r by shift register SR₉. An additional delay in SR₉ is provided to equalize the delay occurring in the time compression and time expansion of the lower subband so that both subbands appear in the output signal s_w in a like time position. The receiving side recovery of the understandable initial signal of the second transmitted signal is accomplished in the same way. The compression and expansion circuits ZK₈ and ZE₉ can basically correspond to the apparatus shown in FIGS. 10 or 9.

Note that the flank steepness of the usual filter is limited; thus, there is, additional (not shown in the drawings) frequency spacing provided in order to prvent disturbing additional signals in the frequency range. On account of the frequency dependent transmission time of the channels used, the edges of the individual signals appear cut-off on occasion, no longer in exactly defined time position, but spilled over with respect to time. It can, therefore, be recommended, e.g. with the time multiplex system of FIG. 15, that the individual signal segments to be spread out, be foreshortened so that they are separated by a sufficiently long pause. These pauses serve for preventing an undesirable temporary cross-talk as a result of time of transit dispersion of the transmitting channels.

A reduction of audible noise is possible through "weak modulation" for forming the reduced segments. This occurs through non-instantaneous changing of the coupling over a short transition time instead of the sudden switching in and out.

For diminishing audible noise and for improving the comprehension, an additional signal treatment can be of use: through the introduction of an additional reverberation on the transmission side, very short speech characteristics which can be fully suppressed through the temporary blending in the upperband become so far dispersed that part of it remains preserved in spite of the screening. Through introduction of an addition reverberation on the receiving side, the joint abutment in the upper subband becomes somewhat equalized.

By the reduction of two or more subbands, the introduction of unequal segment lengths in the different subbands can be recommended because the duration of significant speech characteristics do not agree in all frequency strata. This step can be of especial use in the transmission of music.

Besides the disclosed embodiments, many additional variations thereof are obvious from the techniques described herein, by which the subband is thinned for the reduction of the bandwidth for improving the utilization of the transmitting channels for multi-channel transmission or also for other uses as redundance reduction. The modern known methods of digital signal manipulation such as filtering, time expansion and compression, and storage can be employed with the present invention.