JPH026247B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to an automatic tuning system (AFC) that automatically matches a transmitting (carrier) frequency and a receiving frequency in voice communication using a carrier suppressed single sideband communication system (hereinafter referred to as SSB). It is something.

In recent years, SSB has been widely used in shortwave communications and very high frequency communications to effectively utilize transmission power and frequency bandwidth. However, when using SSB, unlike FM, a carrier wave is not sent, so AFC cannot be applied to achieve stable reception. Also, unlike AM and FM, the difference between the transmitting frequency and the receiving frequency directly causes distortion of the received audio, so detuning causes a significant deterioration in quality, so it is extremely important to maintain synchronization. Therefore, communication is normally performed by carefully tuning the receiver so that the quality of the received voice is at its best. In this case, the accuracy of tuning is said to be around 30Hz, even if you are very skilled. Further, even once the device is tuned, the transmission frequency and the reception frequency gradually change, so it is necessary to re-tune the device over time.

By the way, in cases where no manual intervention is required during reception, or in cases where the automatically received audio is recorded and used later, it is necessary to listen to it as it is, even if the detuning becomes large and the quality deteriorates. . The SSB audio received after being detuned will be re-tuned to SSB.
There is no other way than to perform modulation and demodulation and make corrections in between. However, this is not an easy task and requires skilled manpower.

Therefore, AFC has also been proposed for SSB.
For example, a method has been considered in which a carrier wave is left in the transmitter, which is detected by a narrowband filter on the receiving side, and the reception frequency is controlled using the output signal as a clue to apply AFC. However, this method cannot be applied if the carrier wave is sufficiently suppressed, and there are various problems in stably detecting the carrier wave using a narrowband filter.

In addition, we have developed a method (U.S. Pat. No. 2,938,114, May 24, 1960) in which the waveform obtained by direct envelope detection of the SSB waveform is compared with the waveform obtained by normal detection, and the receiving frequency is controlled by a type of phase synchronization. , a similar method has been proposed. However, these have problems both in principle and technically, and are currently not practical.
There is no AFC for SSB.

The present invention focuses on the fact that voiced sound has a harmonic structure with respect to its fundamental frequency, detects the difference between the transmitting frequency and the receiving frequency from the cepstrum of the received voice, and automatically controls the receiving frequency of the receiver. It is characterized by synchronization, and its purpose is to introduce AFC into the SSB receiver and perform stable SSB communication.

First, the short-time frequency spectrum when audio is transmitted using SSB will be explained with reference to FIG. In Figure 1,
1 is the short-time frequency spectrum of the original voice, 2 and 3 are the short-time frequency spectra when 1 is received after being detuned by F, and the amplitude is shown as an amplitude spectrum expressed logarithmically. Here, 2 is F<0, and 3 is F>0, and when completely tuned, a signal with the same amplitude spectrum as 1 is obtained. Note that the positive/negative of F and the vertical relationship between the transmitting and receiving frequencies differ depending on the rise and fall of the SSB sideband waves, but here, the positive/negative of F will be considered based on the relative relationship between the spectrum of the original voice and the received spectrum. Speech can be divided into voiced sounds and unvoiced sounds depending on how it is generated. FIG. 1 shows an example of voiced sounds, and voiced sounds will be explained below.

By the way, a voiced sound has a waveform caused by the vibration of the vocal cords as its sound source, and is a sound wave that resonates in the vocal tract and is radiated into the air. Since the vibration of the vocal cords is quasi-periodic, speech has a quasi-periodic waveform with the vibration frequency of the vocal cords as the fundamental frequency fo, and its short-time frequency spectrum has a harmonic structure consisting of the fundamental frequency and its harmonics. ing. 1, the local peaks on the amplitude spectrum are harmonically related. On the other hand, in the case of 2 and 3, there is no harmonic relationship even if a peak similar to that of 1 is observed. Speech with a distorted harmonic structure not only gives a sense of discomfort, but also degrades intelligibility when F is large.

The embodiment of the first invention is shown in the block diagram of FIG.
This will be explained with reference to the processing flowchart in FIG. 3, the amplitude spectrum diagram in FIG. 4, and the cepstrum diagram in FIG. 5. Here, FIG. 2 is a block diagram showing the configuration of a receiver that implements the method of the present invention, where 8 is an SSB receiver and 9 is a detuned frequency measuring section. FIG. 3 shows the flow of processing performed at 9. 10 is an audio input terminal and an output terminal of the SSB receiver, and 20 is a receiving frequency control signal output terminal. Also, 4 in Figure 4,
5 is a rewrite of 2 and 3.

The SSB audio waveform signal received at 8 is added to 10 in FIG. This waveform has a section length (window length) T
, and Fourier transform is performed to convert it into a short-time frequency spectrum G(f). The amplitude spectrum log|G(f)| converted into a logarithm is 4 when the detuning frequency F is positive and 5 when it is negative. When this amplitude spectrum is Fourier-transformed again, the cepstrum shown in FIG. 5 is obtained. By the way, in the cepstrum of a voiced sound, a remarkable maximum value is observed in the quefrency corresponding to the fundamental period (1/fo) of the voice. The method of determining the fundamental period from this maximum value is widely used as the so-called "cepstrum pitch extraction method." By the way, even in the case where there is no harmonic structure as in 4 and 5, if the fluctuation of the amplitude spectrum is periodic, the cepstrum that is the result of frequency analysis will have a maximum value corresponding to the fundamental period. However, the value of its local maximum (assumed to be M)
is large when F is 0 or when F is equal to a harmonic of half the fundamental frequency (f ₀ /2), and is small in other cases. This is because, in cases such as 4 and 5, when the amplitude spectrum is viewed as a periodic wave, the phases of the waveforms are different on the positive side and negative side of the frequency axis. On the other hand, when F is 0, the amplitude spectrum is periodic across the positive and negative sides of the frequency axis, and M becomes large. Utilizing these properties, the unknown F is approximately estimated.

Shift f of log|G(f)| by h and calculate the cepstrum as f−h. Change h one after another,
Find the relationship between h and M. Among them, if h is hp when M is maximum, this is the frequency corresponding to F. Note that the range of movement of h may normally be within ±(about f ₀ /2), and may be fixed at about ±50 to ±100 Hz, considering that tuning is initially performed manually upon reception.

The reception frequency of 8 is controlled by the hp obtained in 20. In this case, 8 must be equipped with a circuit that changes the reception frequency using voltage or a synthesizer that can be controlled externally. It is sufficient to generate a required voltage or digital signal in the control signal generating section, and there are no restrictions on the method.

By the way, the direction of control of the receiving frequency differs depending on the upper and lower sides of the SSB sideband. For example, the receiving frequency must be fr+hp for upper sideband waves and fr−hp for lower sideband waves. When tuning manually or when the communication partner is known, this control is easy because the upper and lower sideband waves are also known. In order to apply AFC without worrying about the rise and fall of sideband waves, the following method can be considered. Constant coefficient k for hp (0<k<1)
8 to control the receiving frequency. When the next HP obtained is smaller than the previous HP, the same control is continued. When it becomes large, control is performed by reversing the sign of HP. If you repeat this process several times, fr
converges to ft.

Although the principle of the present invention has been explained above in an analog manner, a supplementary explanation will be given of an embodiment using digital processing. Assume that the received signal applied to 10 is sampled at 10kHz. T=51.2ms (number of samples
512 points), convert it to a short-time frequency spectrum using fast Fourier transform, and then repeat the conversion to a cepstrum to find hp. In this case, since we used a 512-point fast Fourier transform,
The frequency step unit of the amplitude spectrum is approximately 20 Hz, and the unit of the movement of h is also equal to this. Therefore, the error between F and hp is at most about half this unit, about 10 Hz. SSB manual tuning accuracy is 30
Considering that it is about Hz, this is sufficient accuracy. Furthermore, even with 256-point fast Fourier transform,
Since accuracy within 20Hz can be obtained, there is no problem in practical use. More accurate tuning is possible if processing is performed with a longer window length, but since audio is a quasi-stationary signal, the 20 to 50 ms used in normal audio information processing is
It is appropriate to process with a window length of

In the first invention, since the accuracy of AFC is within 10 to 20 Hz, it is not satisfactory when more accurate tuning is required. Therefore, a more accurate embodiment of the second invention will be described with reference to the processing flowchart in FIG. 6, the amplitude spectrum in FIG. 7, and the cross-correlation function diagram in FIG. 8. Note that in the system of the invention shown in FIG. 2, the configuration of the receiver is the same as that shown in FIG.

In Fig. 6, the SSB audio waveform signal added to 10 is cut out with section length T, and then subjected to Fourier transformation to
Convert to (f). This amplitude is converted into a logarithm to calculate the cepstrum, and the process up to this point is the same as the process shown in FIG.

Next, the fundamental period (1/f ₀ ) is determined from the quefrency indicating the maximum value on the cepstrum. This process is the same as when cepstrum is used in the so-called "pitch extraction method." Once f ₀ is determined, the frequency of f ₀ is set as the frequency of the fundamental wave, and a spectrum of constant amplitude consisting of its harmonics is generated. FIG. 7 shows this amplitude spectrum log|S(f)|. Next, with the frequency difference h between both spectra as a variable, log
Calculate the cross-correlation function R(h) of |G(f)| and log|S(f)|. This is shown in FIG. The frequency hp of the difference between both spectra can be easily determined from the maximum value. The processing after determining hp is the same as in the first invention.

As in the first invention, the accuracy of this method will be considered in the case of digital processing. When f ₀ is accurately determined, the accuracy of this method is dominated by the order of harmonics included in G(f) and the frequency resolution of the amplitude spectrum. If G(f) is calculated in 20Hz increments and up to the Nth harmonic clearly exists, hp can be calculated with an accuracy of approximately (10/N) Hz. N
= 10, an accuracy of about 1 Hz can be expected. In addition,
Unlike the first invention, the cross-correlation function can be calculated by setting h in small increments (for example, 1 Hz). Regarding accurate determination of f ₀ , various methods have been proposed in the "cepstrum pitch extraction method" such as the inner loop method and the method using moment calculation, so these can be used.

The process of determining hp is possible when the speech is voiced and quasi-stationary. This process cannot proceed in the case of silence, unvoiced sound, or transient sound, but this can be easily determined from the value of M.
Also, if you do not want to calculate HP, you can just use the known HP as is. Also, since the changes in the SSB transmission frequency and reception frequency are usually not so rapid, it is possible to significantly skip the process of calculating hp to reduce the computer load, use a slow computer, or It is also possible to control the reception frequencies of a plurality of receivers with one 9.

In principle, the method of the present invention may make an error when F and f ₀ or harmonics of f ₀ match. However, such a probability is small, and it is not a problem because it is not spoken at a constant f ₀ over a long period of time.

As described above, according to the method of the present invention, it is possible to automatically synchronize with SSB, so it can be widely used not only for communication between fixed stations but also for mobile radios, fishing radios, etc. to improve call quality. In addition, the labor of the receiver operator can be reduced. Additionally, by measuring the reception frequency, it is possible to accurately determine the SSB transmission frequency, which is normally difficult to measure, which can be useful for establishing orderly communications and monitoring radio waves.

[Brief explanation of drawings]

Figures 1, 4, and 7 are amplitude spectra, Figure 2 is a block diagram, Figures 3 and 6 are processing flow diagrams, Figure 5 is a cepstrum, and Figure 8 is a diagram of a cross-correlation function. be. 1...Original audio (or correctly received audio),
2 and 4...Sound received with detuning (detuned frequency F<0), 3 and 5...Same as above (F>0).

Claims (1)

[Claims] 1 Carrier suppression single sideband communication system (hereinafter referred to as SSB)
In the receiver, a waveform signal of a certain section length of the received audio signal is extracted, converted to a short-time frequency spectrum G(f) by Fourier transform, its cepstrum is calculated, and the peak value corresponding to the fundamental period on the cepstrum is calculated. Find M, then move f by h and calculate the cepstrum as f-h to find M, then h
By repeating the process of calculating the cepstrum by changing
SSB by controlling the receiving frequency of the receiver by HP
Characterized by automatic receiver tuning
SSB automatic tuning method. 2 In the SSB receiver, a waveform signal of a certain section length of the received audio signal is extracted and converted into a short-time frequency spectrum G(f) by Fourier transformation, its cepstrum is calculated, and the fundamental period (1/f) is calculated from the cepstrum. ₀ ), then calculate the amplitude spectrum consisting of harmonics of f ₀ and the cross-correlation function of log|G(f)|, and calculate the frequency hp of the difference between the two spectra from the frequency showing the maximum value of the cross-correlation function The SSB automatic tuning method is characterized in that the reception frequency of the receiver is controlled by HP, and the SSB receiver is automatically tuned. 3 Obtain the frequency hp of the difference between the transmitting frequency and the receiving frequency, and control the receiving frequency of the receiver using hp
In the SSB receiver, k·hp (where 0<k<
By changing the reception frequency by 1), and when the next hp obtained is smaller than the previous hp, control is continued as is, and when it is larger, the sign of hp is reversed and control is performed, thereby reducing the sideband wave of the SSB signal. The SSB automatic tuning method is characterized by automatic tuning without considering the top and bottom.