GB2035761A - Improvements in processing of signals of complex waveform - Google Patents

Abstract

A method and apparatus for modifying the frequency of a complex waveform comprises dividing the frequency spectrum of the signal, or at least a significant part of the said frequency spectrum, into a plurality of relatively narrow frequency bands, passing each, or at least some of, said frequency bands to corresponding frequency multiplying means, combining the outputs and then recording the modified signal at one speed and reproducing it at a slower speed to provide a reconstruction, or at least an approximate reconstruction, of the original signal at a different frequency.

Description

SPECIFICATION inprovements in processing of signals of complex waveform This invention concerns improvements in and relat ing to the processing, more especially frequency modification, ofsignals of complex waveform Signals of complex waveform, for example human.

speech signals which include tones and harmonics extending over a broad frequency spectrum,are extremely difficult to process with a view-ta madification af the frequency af the signal For example, such signals cannot be changed in frequency-by a simple frequency multiplier or divider such as can he applied to a simple sine wave signal The variation in frequency of a complex waveform such as a human speech signal is, however, a function which is highly desirable and would prove extremely useful. in various applications.

Forexample, in cinematography, motion afan action can be slowed down to enable the study of the action in slow motion. No corresponding facility at present exists in relation to study of a passage af human speech, whether or not it accompanies a cinematographic film. Such a facility would be extremely useful, for example in the learning of languages, or atherwise understanding speech which is too fast for the listener do follow.Although limited attempts have been made ta improve the intelligibility of recorded speech by, for example, slowing down the speed af a tape recorder, there is a limit to the utilisation af such a technique, since with the reduced speed of the tape the frequencies afthe various components of the speech waveform are reduced two the extent thatthe speech signal becam- es unrecagnisable..

It is accordingly an object ofthe present invention.

to provide means enabling the modification of the frequency of a complex waveform with a view to subsequent processing of the waveform without a reduction in intelligibility.

In accordance with the invention there is provided a method for modifying the frequency af a complex waveform, which comprises dividing the frequency spectrum af the signal, ar at least a significant. part of the said frequency spectrum,. into a pl u ral ity af relatively narrow frequency- bands, passing- each, ar at least some of, said frequency bands to corres- panding frequency modifying means, and then combining the outputs of said frequency modifying means ta provide a reconstruction, or at least an approximate reconstruction, af the original signal at a different frequency.The said frequency modifying means may comprise frequency multipliers, as this in practice may prove.the preferable frequency modifying function, butthe method of the invention may equally be applied ta frequency division.

The invention furtherprovides an apparatus for modifying the frequency af a complex signal wave- form, comprising a plurality af narrow bandpass filters each arranged to pass frequencies of a predetermined, different relatively narrow frequency band, the said bandpass filters passing frequency bands which are either contiguous arspaced irr a regular or other predetermined manner over a predetermined, broader, frequency spectrum, means for providing a common: signal; of complex waveform to the inputs of the respective bandpass filters, frequency modifying means each connected to, the output of a corresponding one af said band pass filters, and means for summing the au.t- puts af each at said frequency modifying means Preferably the said frequency modifying means comprise frequency multipliers, for example frequency doublers.

By util isation of the method andl means provided in accordance with the invention, it is possible, for example, to mu Itiplythefrequency of a complex waveform, such as the signal waveform of a hiirnan voice,wfthout altering the timing of consecutive signal components, for example corresponding to spoken. syllables of a speech message. This enables the- spoken message to be recorded on'. a medium such. as a moving tape, after multiplication: of its frequency.By then replaying the tape at a propor tianately lower speed determined by the factor by which. the frequency of the angina signal: was multiplied, it is then possible to reproduce a speech signal having the original intonations af the spoken message, but which the spoken. syllables are reproduced more slowly, and therefore more intelligibly, than in the original spoken message The invention is illustrated by way of example in the accompanying diagrammatic drawing, which shows ablock diagramm of one embodiment of apparatus in accordance with the- invention Referring to the drawing, the autputfmm. a microphone 1 is applied via an amplifier 2 to the repective inputs of a plurality of bandpass filters 3 to 7 each of which has a relatively narrow frequency band and the respective frequency bands of which are spaced. over an audible spectrum appropriate to the intelligibility ofthe human voice signal. The outputs from the bandpass filters are each con- nected to the input af a corresponding frequency change 8 to 12, for examples a frequency doubler.

The respective outputs from the frequency changes 8 to 12 are connected in comman to the input of a summing amplifier 13, at output of which is connected to a. tape recorder 14.

When a spoken signal: is transmitted via the microphone 1 and amplifier 2 to the inputs of the respective bandpass filters 3 to 7, the latter serve to analyse the voice signal into relatively narrow frequency components, each of which is capable of being changed in frequency by the respective frequency changer. The width at the frequency band capable af being transmitted by' ea bandpassfilter necessarily corresponds to that which is capable of being handled by the frequency changer which my be of any know type. The number of bandpass filters and their distribution over the audible range of the frequency spectrum. is selected with a view to the optimum analysis and reconstruction of the audible voice signal. The techniques for such voice analysis; are well known in relation to so-called "vocoders" and need not to be described in further detail. in a typical example, however, a total of possibly 20 bandpass filters might be spaced apart overs a frequency spectrum of approximately 10 kilohertzs.

The outputs from the frequency changers are fed to the input of a summing amplifier 13, and thence to a tape recorder 14, by means of which the voice signal may be recorded.

Assuming that the frequency changers 8 to 12 are frequency doublers, the tape recorder 14 will be operated, during recording to the voice signal, at a speed approximately twice that of the normal recording speed, and for subsequent reproduction of the voice signal will be operated at its normal reproduction speed, i.e. half the speed of recording, whereby the originally recorded voice signal is reproduced with the individual components of the signal, i.e. the spoken syllables, reproduced at half the original speed. I However, in view of the modifica- tions of the frequency of the spoken signal, the original intonation and voicing of the respective speech syllables will be maintained, and thus an intelligible spoken signal will be reproduced at half its original spoken speed.

It will be understood that such a process presents a considerable advantage, for example for students wishing to learn a foreign language utilising an example of speech as spoken by a native speaker of the language whose natural speed of talking exceeds the capability of the listener for distinguishing the spoken words. It will be appreciated that although in the example illustrated the frequency multiplication and thus the slowing of the original talking speed is a factor of two, any desired factor of frequency change may be employed as is convenient for the intelligible reproduction of the original spoken message.

Claims (4)

1. A method for modifying the frequency of a complex waveform, which comprises dividing the frequency spectrum of the signal, or at least a significant part of the said frequency spectrum, into a plurality of relatively narrow frequency bands, passing each, or at lease some of, said frequency bands to corresponding frequency modifying means, and then combining the outputs of said frequency modifying means to provide a reconstruction, or at least an approximate reconstruction, of the original signal at a different frequency.

2. An apparatus for modifying the frequency of a complex signal waveform, comprising a plurality of narrow bandpass filters each arranged to pass frequencies of a predetermined, different relatively narrow frequency band, the said bandpass filters passing frequency bands which are either contiguous or spaced in a regular or other predetermined manner over a predetermined, broader, frequency spectrum, means for providing a common signal of complex waveform to the inputs of the respective bandpass filters, frequency modifying means each connected to the output of a corresponding one of said bandpass filters, and means for summing the outputs of each of said frequency modifying means.

3. An apparatus as claimed in Claim 2, wherein the said frequency modifying means comprise frequency multipliers.

4. An apparatus for processing speech signals, substantially as described herein with reference to the accompanying drawings.

4. A process for modifying signals of complex waveform, substantially as described herein with reference to the accompanying drawing.

5. An apparatus for modifying signals of complex waveform, substantially as described herein with reference to the accompanying drawing.

New claims or amendments to claims filed on 16 January 1980 Superseded claims All (1 to 5 inc.) New or amended claims,:-

1. A method for processing speech to improve the intelligibility thereof including the steps of dividing the frequency spectrum of a speech signal, or at least a significant part of the said frequency spectrum, into a plurality of frequency bands, passing each, or at least some of, said frequency bands to corresponding means for multiplying the frequency of signals within the respective frequency bands, combining the outputs of the said frequency multiplying means to provide a synthesis of the original speech signal in which the speech tones are reproduced at a higher pitch, recording the synthesised signal by means of a recording means operating at a first speed, and then reproducing the recorded signal by means of a reproducing means operating at a second speed, lower than the first, whereby there is obtained a reconstruction of the original speech signal with the speech tones reproduced at an intelligible pitch and with the spoken words reproduced at a slower speed than in the original signal.

2. An apparatus adapted to carry outthe method of Claim 1, said apparatus comprising a plurality of bandpass filters which respectively pass signals within different frequency bands, all of said bandpass filters being connected in parallel to a common signal input, a plurality of frequency multipliers each having an input connected to the output of a corresponding one of said bandpass filters and all having the same multiplication factor, the outputs of all of said frequency multipliers being connected in parallel to a common summation means, and a sound recording and reproducing apparatus to a recording input of which the said summation means is connected, the said recording and reproducing means being arranged to record at a first, relatively high speed and to reproduce at a second relatively low speed, and the arrangement being such that the ratio between the said recording and reproducing speeds of said recording and reproducing means substantially corresponds to the multiplication factor of said frequency multiplying means.

3. A method for modifying speech signals, substantially as described herein with reference to the accompanying drawings.