Bandwidth compression eliminating frequency transposition and overcoming phase ambiguity
    32.
    发明授权
    Bandwidth compression eliminating frequency transposition and overcoming phase ambiguity 失效
    带宽压缩消除频率传输和覆盖相位漂移

    公开(公告)号:US3484556A

    公开(公告)日:1969-12-16

    申请号:US3484556D

    申请日:1966-11-01

    CPC classification number: G10L25/00

    Abstract: 1,188,014. Bandwidth compression. WESTERN ELECTRIC CO. Inc. 11 Oct., 1967 [1 Nov., 1966], No. 46392/67. Heading H4R. In a bandwidth compression system the speech signal is split into a number of contiguous frequency bands to provide a subsignal S(t) which is treated as the real part of an analytic signal, # (t) = s(t) + j###S(t) where #S (t) is the Hilbert transform of s (t), and the signal which is transmitted is the real part of the square root of the analytic signal which is shown to be where a (t) is the instantaneous amplitude of the analytic signal # (t), and # (t) = a (t) ej# (t). The signal transmitted is limited in frequency band to half the input frequency band and is a signal from which the original frequency band may be synthesized. As shown in Fig. 1 the input speech band is split into contiguous frequency bands by filters 100, having a bandwidth such that no filter contains more than one formant. The resulting channel signals are applied to Hilbert transform networks 102, comprising appropriately adjusted transversal equalizers, and to a delay 101 of value equal to the delay imposed on the signal in network 102 to produce appropriately synchronous value s (t) and #S (t) from which, by squaring, adding, and square rooting, in 103, 104, 105 and 106 the instantaneous amplitude of the analytic function is derived. The resulting signal, a (t), is added to the original signal, s (t), in 107, multiplied by ¢ in 108, and the square root taken in order to obtain the real part of the square root of the analytic signal. To resolve the ambiguity introduced by the square rooting a switch 111 is arranged to invert the output of 110 every time signal s (t) goes through zero when the signal s (t) is positive using the circuit described with respect to Fig. 3 (not shown). It is shown that the original signal s (t) may be obtained from the transmitted real part of the square root of the analytic function simply by producing the imaginary part of the square root in Hilbert transform network 114, delaying the real part in 113 to synchronize it with the imaginary part, squaring both parts in 115 and 116 and taking the difference in 117. The resulting signals from the similar channels 1 to N are added in 10 and reproduced at 11. In a modification Fig. 4 (not shown) the amplitude of the real part of the analytic signal is restored to its original amplitude, a (t), prior to transmission by multiplying the output of the square root network 110 by the output of a similar square root network connected to the output of square root network 206. At the synthesizer the output of the subtractor 117, which is now a (t) s (t), is divided by a signal, a (t), derived by adding the outputs of networks 115 and 116 and taking the square root. The arrangement is said to produce instantaneous frequency division.

    Time domain vocoder
    37.
    发明授权
    Time domain vocoder 失效
    时域声码器

    公开(公告)号:US3071652A

    公开(公告)日:1963-01-01

    申请号:US81202859

    申请日:1959-05-08

    CPC classification number: G10L25/00 G10L19/02

    Abstract: 954,487. Telephone transmission systems. WESTERN ELECTRIC CO. Inc. April 28, 1960 [May 8, 1959], No. 14877/60. Heading H4R. Information relating to a message wave, e.g. a speech wave, is obtained by deriving from the message wave, a secondary wave coherent with the periodicities of the message wave but having a flat frequency spectrum, and determining the cross-correlation between these signals for various relative lags between the two waves. Signals representative of these correlations are transmitted over narrow band channels to enable synthesis of the message wave to take place at a receiver. In one embodiment the secondary wave is obtained by heavy clipping of the message wave. In a second embodiment the abovementioned clipper is succeeded by a differentiating circuit and a rectifier, so that the secondary wave consists of a pulse each time the message wave crosses the zero line in a particular direction. In a preferred form, however, the secondary wave consists of pulses coincident with peaks of the message wave. These pulses are preferably of amplitude proportional to the peak value of the message wave, or may take on more exaggerated values, effected by means of a volume expander. Thus, as shown in Fig. 3, the message signal is fed directly to a delay line 3 (terminated at 4 for no reflection) and also to a differentiator 11 followed by a clipper 12. The clipper output has a downward - going discontinuity corresponding to each positive peak of the message wave, and a pulse may be derived therefrom either by a differentiator circuit followed by a rectifier, or, as shown, by a monostable multivibrator 13. The output of circuit 13 may be fed direct to the multipliers 15 or preferably may be used to sample the message wave (expanded at 16) the output of the sampler 14 after being subjected to A.G.C. being taken for the secondary wave and fed to a series of multipliers 15, one of which is provided for each tapping of the delay line, and which serves to measure the correlation between the secondary wave and the output from the corresponding tap of the delay line. The outputs of the multipliers feed low-pass filters of narrow band-width, whose outputs #(#) are then transmitted. A pitch signal derived from an 80- 350 c.p.s. band filter is also transmitted. It is stated that in contrast with the earlier described methods, this last method results in a correlation function which is substantially symmetrical about #=0 (# representing the relative delays of the message wave and the secondary wave) and that therefore the number of tappings on the delay line may be halved as compared with the previous cases. In this case, in the receiver, Fig. 5, two signals require to be reconstructed from each one of the signals #(#) received. Thus the pitch signal after passing through a delay equalizer is clipped and each upward change of the clipped signal operates a mono-stable multivibrator 33 to produce an output pulse which is modulated in multipliers 35 by the respective signals #(#). The modulated pulses are fed to appropriate taps on a delay line 40, one of whose ends is open-circuited to provide an in-phase reflection. Thus for each input to a delay line tap # n , two pulses arrive at the load resistor 41, having delays # N Œ# n respectively. The aggregate of these pulses pass through a low-pass filter 43 to produce the required message wave output. In the absence of a fundamental frequency relay 34 falls back to connect a noise source 36 to the multipliers. In this output the relative phases of the various components may be changed as compared with the original wave, but in the case of speech this has little effect on the intelligibility. In a modification, Fig. 4, the correlation is effected by feeding the signal to a high-loss attenuation pad 21 and thence to the delay line 3. Each tapping is then fed to summation circuit 22 and a difference circuit 23 whose other inputs are fed with the original signal. The outputs of these circuits are applied to full-wave rectifiers 24, 25 whose outputs are thus the absolute values of the sums and differences and the difference between these absolute values is taken at 26 and fed to a low-pass filter 27 whose output provides the appropriate one of the output signals #(#). The difference circuit 23 may be omitted, the circuit 26 taking the difference between the output of circuit 24 and the absolute value of the message signal. In all cases the message signal may be differentiated once or twice before being applied to the analysing apparatus, corresponding integrations being performed on the output from the synthesizer at the receiver. Specification 466,327 is referred to.

Patent Agency Ranking