DE1258471B - Channel vocoder for bandwidth-limited message transmission - Google Patents

Description

FEDERAL REPUBLIC OF GERMANY

GERMAN

PATENT OFFICE

EDITORIAL

Int. CL:

H 04b

H04m

German class: 21 a2 - 36/22

Number: 1258 471

File number: J 29098 VIII a / 21 a2

Filing date: October 1, 1965

Open date: January 11, 1968

The invention relates to a channel vocoder for bandwidth-limited message transmission.

The analysis and synthesis of speech signals with the channel vocoder for the purpose of bandwidth-limited message transmission or message processing is known to consist of an analysis and synthesis of the spectral envelope and an analysis and synthesis of the spectral fine structure. The starting point for the invention is that shown in FIG. 1 known channel vocoder shown schematically. The speech signal s (t) is broken down into the _{spectrum channels UC 1} to K _n with the help of the spectrum analyzer SAn and into the excitation _{channel UC 0} by means of the fundamental wave analyzer AAn . The spectrum analyzer S! / 4n consists of a number of band filters BP ₁ to BP _n , which split the speech signal into the desired frequency bands. _{The rectifiers G 1} to G _n and low-pass filters LP ₁ to LP _n following each band filter determine the instantaneous value of the energy present in the individual frequency bands. In the fundamental wave analyzer AAn , a suitable description of the excitation function, which is made available via the excitation channel K ₀ , is obtained from a frequency band ranging approximately from 0 to 300 Hz. With the help of a transmission link Ü, which is implemented in conventional technology and can also include systems for time division multiplex transmission, for example, the signals of the channels K ₀ to K _{n are} transmitted from the transmitting side of the vocoder, where the speech analysis takes place, to the receiving side, where the speech synthesis takes place is carried out. On the receiving side, synthesizers SSy and ASy are assigned to both the spectrum channels and the excitation channel. The spectrum channels K ₁ to K _n have their own modulators M on the receiving side, which modulate the portion of the spectrum signal transmitted in the respective channel with the fundamental wave signal α (t) of the fundamental wave synthesizer ASy. The output signals of all spectrum channels are then passed to a set of band filters BP ₁ to BP _n , at the output of which the synthetic speech signal g (t) is produced.

As already mentioned, the analysis and synthesis of speech signals with the channel vocoder consists of an analysis and synthesis of the spectral envelope and the spectral fine structure. The terms of the spectral envelope and the spectral fine structure are shown in FIGS. 2 and 3 illustrated. For voiced sections of the speech signal, the fine structure F is a line structure due to the quasi-periodic character of the speech signal (FIG. 2). For unvoiced sections of the speech signal, the channel vocoder is used for bandwidth-limited
Messaging

Applicant:

International Business Machines Corporation,

Armonk, N. Y. (V. St. A.)

Representative:

Dipl.-Ing. H. E. Böhmer, patent attorney,

7030 Boeblingen, Sindelfinger Str. 49

Named as inventor:
Dipl.-Ing. Dr. Kurt Bandat,
Dipl.-Ing. Erwin Paulus,
Dipl.-Ing. Dr. Ernst Rothauser, Vienna

Claimed priority:

Austria from October 1, 1964 (A 8384/64)

on the other hand, the power spectrum is continuous (Fig. 3). The information about the spectral envelope E required for the synthesis is provided by the spectrum analyzer SAn in the form of a number of variable direct voltages on the spectrum channels K ₁ to K _n . The information about the spectral fine structure for the synthesis of the excitation function is supplied by a special part of the analyzer ^ 4 ^ 4n. In conventional vocoder systems, it consists of a signal that can only have two values and indicates whether the excitation is to be voiced or unvoiced at the moment, and for voiced excitation additionally in a signal that provides information about the current fundamental frequency. In the synthesis, the information about the spectral fine structure is used in such a way that for voiced sections the excitation function is a pulse sequence with the fundamental frequency determined during the analysis, while for unvoiced sections it consists of noise.

The synthesis of a speech signal with the channel vocoder can be expressed as a multiplication of the short-term power spectrum the excitation function with the spectral determined approximately in the spectral analysis Enveloping are understood. If the spectral envelope of the excitation function is flat, then has the synthetic speech signal has approximately the same spectral envelope as the original speech

709 718/315

bl 258471

3 4

signal. The spectral fine structure is due to the in the individual spectrum channels between specs mentioned properties of the excitation radio spectrum analyzer and transmission link inclusion are also the same as in the original language tet.

signal. The arrangement according to the invention offers against

Even in impulse-excited vocoder systems there are several precedents over conventional vocoder systems Synthesis from the multiplication of the spectral parts:

Envelope with the short-term power spectrum of the Zum first is compensated for by the

Excitation function. The excitation function is the flow of the fundamental speech wave onto the spectrum channels probably for voiced as well as for voiceless tunings - the quality of the synthetic speech thereby blended of the speech signal a series of short io improves the distortion of the spectral envelope pulses. The distance between two successive- ly of the synthetic speech signal, which is determined by successive Pulses is controlled by the analyzer. deviations of the spectral envelope of the an-for voiced sections it corresponds to the period excitation function of the flat envelope duration, for voiceless sections it fluctuates inwardly becoming, being balanced, half of certain limits statistically. The second advantage of this is a reduction in the flow of data The type of excitation is that a synthesis is brought about in the spectrum channels of mixed forms between voiced and unvoiced those in the spectrum channels are unnecessarily possible and that in the synthesis of voiced information about the fundamental speech wave, Sections of the speech signal also have their own channel ready for small changes for their transmission Period duration, which is established from period to period, 20 which is destroyed before transmission, can occur. In the following the invention is based on one

The synthetic language gains in naturalness. An exemplary embodiment illustrated by drawings Disadvantage of this type of excitation results from the described in more detail. It shows

The fact that the spectral envelope of the short-term Fig. 4 is the block diagram of a channel vocoder

power spectrum of the excitation function not to 25 with spectrum compensation,

at any point in time over the entire audio frequency F i g. 5 the block diagram of a controllable lowering area is flat. Deviations from the flat attenuator and

envelopes lead to an envelope of the short-term F i g. 6 the basic circuit diagram of a logarithmic

power spectrum of the synthetic speech signal, circuit.

which compared to that of the original speech signal with- 30 F i g. 4 shows the block diagram of a konvenunter is impermissibly distorted. tional impulse-excited channel vocoder, which

Since the distortions in the sound quality are noticeable by an additional excitation signal generator of _{synthetic speech, it is the ASy 1} , a further spectrum analyzer _{S ^ n 2} and the task of the invention to add a vocoder arrangement through a set of controllable attenuators, R ₁ , which are suitable is, distortions of the spectra 35 to R _{n is} added. The structure of the additional specular envelope of the synthetic speech signal spectrum analyzer SAn ₂ is largely identical to compensate for the flat envelope caused by deviations in the spectral structure of the spectrum analyzer SAn ₁ , which is used in the envelope of the excitation function from that of the conventional vocoder system. The same applies to the structure of the additional attachment

Furthermore, since the output signal is sent to the η parallel 40 excitation signal _{generator ASy 1} . The controllable outputs of the spectrum analyzer SAn not only weaker also have conventional information about the spectral envelope developed, and their control is carried out in such a way that they hold, but rather also information about their attenuation depending on the size of the fundamental wave, is this Signal can be changed in the spectrum channels of the control signal, since a separate channel K _{0 is made} available for the transmission of the fundamental wave information, and then leads to an improvement in the fluid. Quality of synthetic speech and also to one

It is therefore a further object of the invention to reduce the flow of information in the transmission by eliminating this redundancy in the path. If the expansion of the transmission channels in front of the receiving end sees a reduction in the data flow and 50, the information flow in the transmission is thus not reduced by a further reduction in the bandwidth of the path; but the additional channel vocoders can be used for this. Excitation synthesizer ASy _{1 is} omitted, the

A channel vocoder is therefore proposed for transmission of messages limited to a hand width of the already existing excitation synthesizer _{, ASy 2} takes over.

in which the signal component of the fundamental speech wave is compensated for in the 55 spectrum channels generated during voiced speech sections. the additional excitation signal generator ASy ₁ one

Such a vocoder can be realized by the impulse sequence which is similar to the impulse sequence which lent that the known arrangement of the Ka- for the excitation of the band filters in the speech vocoder is expanded in such a way that is used in the synthesis. This pulse train e is connected to spectrum channels with controllable attenuators, also to the additional spectrum analyzer _{SAn 2} , the control signals of which carry the output signals. The output signal e _t of each channel of an additional spectrum analyzer, which of the additional spectrum analyzer is used for control as the input signal, the output signal of a fundamental wave analyzer additionally switched on between the attenuation of the output signals in the corresponding fundamental wave analyzer and additional spectrum channels of the channel vocoder analyzer. The signals of the channels K ₀ to K _n are fed to the synthesizer. given on a transmission link Ü that this

Further favorable results are achieved by sending signals to the receiving end in a suitable manner Order achieved in which the adjustable attenuator of the vocoder transmits. On the reception side of the

Vocoder, the speech synthesis takes place in that the output signals of the excitation signal generator ASy _{2 act} on the band filter of the spectrum synthesizer SSy in such a way that the synthetic speech signal g (t) is produced, which corresponds to a certain approximation to the original speech signal s (t).

F i g. 5 now shows a circuit arrangement known per se which fulfills the function of the required controllable attenuator. According to this, the output signal U _{2 of} the arrangement should be approximately equal to the quotient formed from the channel voltage U ₁ and the control voltage e

= C

Two circuits with logarithmic characteristics 10g ₁ and log _{2 are} provided, which generate the logarithm of the input variables U ₁ and e. The output signals of these circuits are subtracted from one another in the subsequent subtraction circuit Sub, which can have a conventional structure, for example. This creates the value log -. In order to form the number of this value, a circuit arrangement is required which has an antilogarithmic characteristic. A known implementation of this characteristic uses a feedback amplifier V with a high gain, in whose feedback loop a circuit log _s with logarithmic characteristics is switched on, which has the same structure as the circuits ZOg ₁ and log ₂ , for example.

F i g. 6 now shows another known principle of how the circuit is constructed with a logarithmic characteristic. It contains two diodes D ₁ , D ₂ and D ₃ , D ₁ , which are connected in a kind of bridge circuit with the resistors /? and -R 'work together. This circuit makes use of the knowledge that the diodes in the forward direction have a voltage-current characteristic in a certain range, which can be approximated by the following equation: E = r log i.

The circuit according to FIG. 6 subtracts the output values of the two logarithmic circles and

forms the logarithm of the quotient -. In the

subsequent delay circuit DL, which can also consist of a feedback amplifier with logarithmic feedback, for example, the number of the quotient is then formed.

Claims (3)

Patent claims: 1. Channel vocoder for bandwidth-limited message transmission, characterized in that the signal component of the fundamental speech wave in the spectrum channels (K ₁ to K _n in FIG. 4) is compensated. 2. Channel vocoder according to claim 1, characterized in that adjustable attenuators (R ₁ to R _n ) are switched on in the spectrum channels (K ₁ to K _n ) , the control signals (e _t to e _n ) of which the output signals of an additional spectrum analyzer (SAn ₂ ) , to which the output signal [a (t)] of a fundamental wave synthesizer (ASy ₁ ) additionally switched on in parallel to the receiver-side fundamental wave synthesizer (ASy ₂ ) is fed as input signal (e) (FIG. 4). 3. Arrangement according to claim 2, characterized in that the controllable attenuators (R ₁ to R _n ) in the individual spectrum channels (K ₁ to .RT _n ) between the spectrum analyzer (SAn ₁ ) and the transmission path (Ü) are switched on (Fig. 4 ). For this purpose 2 sheets of drawings 709 718/315 12.67 © Bundesdruckerei Berlin