JP2670541B2 - Training signal detection method - Google Patents

Description

Description: BACKGROUND OF THE INVENTION (Industrial application field) The present invention reduces the amount of calculation and storage capacity of an arithmetic processing unit required when detecting a training signal transmitted by a modem, and furthermore, detects a signal. The present invention relates to a training signal detection method capable of improving the certainty of. (Prior Art) As a communication system for transmitting and receiving signals by appropriately switching between voice communication such as a telephone and communication via a modem (modulator / demodulator) such as a facsimile, as shown in FIG. 2, for example. Telephone 1 and modem 2
While connecting in parallel with the facsimile 3 equipped with
A switching circuit 6 selectively switches to either a voice mode or a facsimile mode when a signal unique to each mode is detected by a detector 5 in a switching device 4. When a modem signal is input when the system is in the voice mode, a training signal appearing at the head of the modem signal is detected and the mode is switched to the modem mode. When a voice signal is input when the system is in the modem mode, Detects the arrival of an audio signal and switches to the audio mode. Among them, the method described below is generally used as a means for detecting the training signal. That is, the spectrum of the training signal output from the modem 2 has three frequency components of f ₁ : 0.5 Khz, f ₂ : 1.7 KHz, and f ₃ : 2.9 KHz, for example, in a 9600 bps modem of CCITT Recommendation V.29. These frequency components f ₁ , f ₂ and f ₃ are assigned to three digital band pass filters (BP
F) After extraction by 5a, integrators 5b connected to each BPF 5a respectively
Calculated power W _1, W _2, W ₃ of each frequency component by the presence of a training signal incoming by comparing by the comparator 5c power Wa obtained by these respective power values integrating the training all signals Is determined and a detection signal is output based on the determination result. Based on this detection signal, the system is switched to the modem mode and signal processing such as band compression suitable for the modem signal is performed. However, in the conventional training signal detection method, in order to reliably extract the three frequency components of the training signal spectrum, it is necessary to configure each BPF with two sets of narrow-band multistage digital filters to enhance the frequency resolution capability. was there. For this reason, the calculation amount of the arithmetic processing unit (CPU) increases, and further, it is necessary for detecting the training signal.
It has the drawback of increasing the storage capacity of the CPU. (Purpose of the Invention) The present invention has been made in view of the above, and enables the detection of a training signal using a DFT (Discrete Fourier Transformation Unit) that does not require much frequency resolution accuracy, thereby allowing the training signal to be detected. It is an object of the present invention to provide a training signal detection method in which the amount of calculation and storage capacity of an arithmetic processing unit required for detection are reduced, and the reliability of signal detection is improved. (Summary of the Invention) In order to achieve this object, according to the present invention, a switching device discriminates between a voice signal by a voice communication means and a modem signal by a data communication means, and a voice mode signal processing for processing the voice signal. A specific frequency component for identifying that the modem signal is a modem signal at the beginning of the modem signal, the switching circuit switching between the modem signal processing and the modem mode signal processing for processing the modem signal, and an analog-digital converter. When the voice signal or the modem signal is input, the detector uses the training signal to discriminate between the voice signal and the modem signal, and the switching circuit uses the result of the discrimination. In the switching device for switching between the two, the detector is replaced with two discrete Fourier transform means. , An integrator and a comparator, and the frequencies f4 and f5 are extracted by the two discrete Fourier transform means, where f4 is not a frequency component of the power spectrum of the training signal and the audio signal F5 is a frequency component of the power spectrum of the training signal, and f5 is selected as a frequency at which the power spectrum of the audio signal is substantially zero, and the power W4 of the f4 is selected as: The electric power W5 of the f5,
The switching is performed under the condition that the total signal power Wb in the audio frequency band is W5 / W4> α, Wb / W5 <β, where α and β are predetermined thresholds of 1 or more. It is characterized in that the signal input to the device is determined to be a training signal indicating the modem. In addition to the above characteristics, the above W5 / W4> α and Wb / W5 <β
In addition to the above condition, Wb> γ, where γ is the minimum value that can be considered as the value of the training signal Wb that should arrive, and the condition that the incoming signal is a training signal is also characterized . Further, in addition to the above characteristics, it is also characterized in that discrete power Fourier transforms relating to the frequencies f4 and f5 are performed a plurality of times, and the respective electric powers W4 and W5 are obtained from the average of the obtained respective values. (Example) Hereinafter, the training signal detection method of the present invention will be described in detail. FIG. 1 (a) shows an embodiment of a detector used for carrying out the method of the present invention and a configuration of a switching device using the detector. The switching device 14 is used as a voice communication means connected in parallel. A detector for detecting a frequency component peculiar to each signal from the voice signal and the modem signal output by the telephone 11 and the facsimile 13 (including the modem 12) as data communication means and discriminating the signal. 15
A switching circuit 16 for selectively switching to either a voice mode or a facsimile mode based on a detection signal from the detector 15; And an analog-digital conversion device (A / D) for converting into a signal. The detector 15 includes DFT (discrete Fourier transform means) 20a and 20b for extracting the power of the frequency f4 characterizing the audio signal and the power of the frequency component f5 (equivalent to f3 described in the related art) constituting the training signal. Integrator 21 that integrates the entire training signal and outputs it as a power signal, and
Comparing the power signals W ₄ and W ₅ output from the DFT 20a and 20b with the power signal Wb output from the integrator 21 to determine whether the signal is a modem signal, and outputting the determination result as a detection signal to the switching circuit 16 Comparison And a container 22. The switching circuit 16 is controlled by the signal from the comparator 22 to guide the signal of the modem or the telephone to the signal processing block in each of the modem mode and the voice mode and to output its output to the external line. FIG. 1 (b) shows the power spectrum of the voice signal and the training signal of the V.29 modem, the average voice power spectrum is shown by the solid line, and the power spectrum of the training signal is shown by the dotted line. As is apparent from this figure, in the audio signal, the power component around 1 KHz is relatively large, but there is no power component of the training signal near this. Next, in the training signal, the power component is located near 2.9 KHz, but the power component of the audio signal near this is very close to zero. Judging from these points, when the power component at 2.9 KHz is large and the power component at 1 KHz is small, it can be determined that the training signal has arrived. In other words, a discrete Fourier transform of the incoming signal is performed for two frequencies f ₄ and f ₅ that characterize the difference between the training signal and the audio signal, and W ₄ and f ₄ corresponding to the resulting f ₄ and f ₅ the ratio of _{W 5 (W 5 / W 4} ) is a certain value α
It is more than, and the incoming signal on the condition that the ratio of the total signal power Wb and W ₅ (Wb / W ₅₎ is below a certain value β can know that it is a training signal. From such a point, in the case of V.29 modem, each DFT 20a and 20
Set the center frequencies of b to 1KHz and 2.9KHz, respectively. Here, α and β are one or more predetermined thresholds. As described above, the method of the present invention is a DFT with a relatively small amount of calculation.
Since the training signal is detected by using, the calculation amount and the storage capacity required for the CPU for detecting the training signal can be significantly reduced. According to the experiment, in the comparison between the conventional example and the method of the present invention, the calculation amount is changed from 96 to 70, and the required storage capacity is reduced.
We were able to reduce from 100 to 18 respectively. Therefore, by using the residual capacity obtained by reducing the used capacity, it is possible to increase the operation monitoring program for other circuits and increase the data retention time for training signal analysis. Can be built. In addition, in addition to the condition of W ₅ / W ₄ > α, Wb / W ₅ <β in the above embodiment, a condition of Wb> γ (γ is a minimum value considered as a value of the training signal Wb to arrive) is further satisfied. It can also be configured to judge that the incoming signal is the training signal only when the condition is satisfied. By adding such a condition, it is possible to prevent erroneous detection due to an unstable noise component included in the spectrum. Further, the discrete Fourier transform in the predetermined frequency f ₄ and f ₅ performed a plurality of times, when to obtain the average from W4, W5 of the values obtained, it is possible to perform determination based on the average spectrum Therefore, accurate detection with few errors can be performed. Incidentally, in implementing the present invention, the above-mentioned detection circuits may be individually provided as independent devices, but may be added as one function of a program by utilizing a CPU incorporated in a required functional block to be applied. Things will not need explanation. In the above embodiments, the case where the DFT is used as the frequency component extracting means is shown, but the DFT includes FET (fast Fourier transform) or DCT (discrete cosine transform) which is a Fourier transform means in a broad sense. It will be obvious that it will be done. (Effect of the Invention) As described above, the method of the present invention detects a training signal using a DFT, which has a relatively small calculation capacity, in place of the narrow-band bandpass filter used in the conventional method. Therefore, the amount of calculation and the storage capacity required for the CPU for detecting the training signal can be significantly reduced, and the detection sensitivity similar to that of the conventional method can be obtained.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 (a) is an explanatory view showing an embodiment of a detector used for carrying out the method of the present invention and the configuration of a switching device using this detector, and FIG. 1 (b). Is a level distribution diagram of a voice signal and a signal, and FIG. 2 is a configuration explanatory diagram of a detection device used in a conventional detection method. 1 ... Telephone, 2 ... Modem 3 ... Facsimile, 4 ... Switching device 5 ... Detector, 6 ... Switching circuit 12 ... Modem (modulator / demodulator) 13 ... Facsimile, 14 ... Switching device, 15 …… Detector, 16 …… Switching circuits 20a, 20b …… DFT (discrete Fourier transform means), 21 …… Integrator

Claims (1)

(57) [Claims] The switching device is a detector that discriminates a voice signal by voice communication means and a modem signal by data communication means, a switching circuit that switches between voice mode signal processing for processing the voice signal and modem mode processing for processing the modem signal, An analog-digital converter is provided, and the modem signal is provided with a training signal having a specific frequency component for identifying that the modem signal is a modem signal, and when the voice signal or the modem signal is input, The detector uses the training signal to discriminate between the audio signal and the modem signal, and uses a result of the discrimination to switch the switching circuit. In the switching device, the detector includes two discrete Fourier transform units. , An integrator, and a comparator. The frequency conversion means extracts frequencies f4 and f5. At this time, f4 is not a frequency component of the power spectrum of the training signal, and is selected as a frequency near a peak value of the power spectrum of the audio signal. A frequency component of the power spectrum of the signal, and a frequency at which the power spectrum of the audio signal is substantially zero, the power W4 of the f4, and the power W5 of the f5,
The switching is performed under the condition that the total signal power Wb in the voice frequency band is W5 / W4> α, Wb / W5 <β, where α and β are predetermined thresholds of 1 or more. A training signal detecting method, comprising: determining that a signal input to the device is a training signal indicating the modem. 2. In addition to the above W5 / W4> α and Wb / W5 <β, Wb>
The training according to claim (1), wherein the arrival signal is determined to be a training signal when the condition γ, where γ is the minimum value considered as the value of the training signal Wb to arrive, is also satisfied. Signal detection method. 3. The training signal detection according to claim 1 or 2, wherein the discrete Fourier transform relating to the frequencies f4 and f5 is performed a plurality of times, and the respective powers W4 and W5 are obtained from an average of the obtained values. Method.