Method and device for the recognition of modulations

• US 5,271,036 A
• Filed: 11/15/1991
• Issued: 12/14/1993
• Est. Priority Date: 11/16/1990
• Status: Expired due to Fees

• Current Assignee: Thomson-CSF
• Original Assignee: Thomson-CSF
• Inventors: Lobert, Bruno, Sourdillat, Bruno
• Primary Examiner(s): Kuntz, Curtis
• Assistant Examiner(s): GHEBRETINSAE, TEMESGHEN
• Application Number:
  US07/792,913
• Time in Patent Office: 760 Days
• Field of Search: 375/10, 370/13, 370/69.1, 455/226.1
• US Class Current: 375/227
• CPC Class Codes: G06F 2218/08 Feature extraction
• CPC Class Codes: G06N 3/045 Combinations of networks
• CPC Class Codes: H04B 14/002 characterised by the use of a carrier modulation using subcarrier modulation H04B14/08
• CPC Class Codes: H04L 27/0012 arrangements for identifying the type of modulation

Abstract

A method for recognizing modulation of radioelectric transmissions from instantaneous spectra of transmission observed in a determined frequency band by a Fast Fourier Transform spectrum analyze is disclosed wherein the method involves the calculation of a plurality of parameters for each transmission spectrum line which is observed. The parameters include the calculation of a mean amplitude of all the lines k of the spectrum contained in the determined frequency band and the calculation of a signal-to-noise ratio as well as a standard deviation in amplitude of each line of the spectrum. Also calculated is a coefficient of correlation COR(k,k) of amplitude of each line k with the homologous lines of the transmission spectrum contained in the determined frequency band. Subsequently, a comparison is made, through a network of neurons, of the parameters of each transmission spectrum with expected transmission parameters and an indication is made as to a correspondence between a transmission and an expected transmission when the difference detected by the comparison is at a minimum value.

Claims

1. A method for the recognition of the modulation of radioelectric transmissions from instantaneous spectra of transmission observed in a determined frequency band by a Fast Fourier Transform spectrum analyzer, said method comprising the steps of, for each transmission spectrum line observed in the determined frequency band:

calculating a mean amplitude parameter of all the line k of the spectra contained in the determined frequency band;

calculating a signal-to-noise ratio parameters RSBk;

calculating a standard deviation parameter ETk in amplitude of each line of the spectrum;

calculating a coefficient of correlation parameter COR(k,k) of each amplitude of each line k with the homologous lines of the transmission spectra contained in the determined frequency band,

said method subsequently further including making a comparison, through a network of neurons, of results of said calculated parameters of each transmission spectrum with expected transmission parameters and indicating a correspondence between a transmission and an expected transmission when the difference detected by said comparison is at minimum values.

2. A method according to claim 1 consisting in associating, in the network of neurons, a neuron with each type of modulation.

3. A method according to claim 2, consisting in declaring that a type of modulation is found when the output level of the associated neuron is the maximum.

4. A method according to claim 3 consisting in determining the output level Oi of a neuron by a relationship of the form:

##EQU4## where Theta _i designates a threshold associated with the neurone i,

X designates an input j,

N_i is the number of inputs of the neuron i,

and Omega_ji designates the weight assigned to the input j.

5. A method according to claim 3, consisting in the determining of the output level O_i of a neuron by a relationship of the form:

##EQU5## where Theta _i designates a threshold associated with the neurone i,

X_j designates an input j,

N_i is the number of inputs of the neuron i, and Omega_ji designates the weight assigned to the input j.

6. A method according to claim 4 consisting of the adjusting, in a learning phase, of the weight Omega_ij of each neuron corresponding to an expected type of transmission in comparing, for each expected transmission, the output of the network with the expected output in retro-propagating the error obtained according to a gradient type of algorithm which minimizes this error.

7. A device for recognizing modulation of radioelectric transmissions from instantaneous spectra of transmissions observed in a determined frequency band by Fast Fourier Transform, comprising:

a receiver coupled to a network of neurons by means of a spectrum analyzer; and

a device for extracting transmission parameters including the mean amplitude of all lines k of the spectra contained in the determined frequency band and a signal-to-noise ratio RSBk, and a standard deviation ETk in amplitude of each line of the spectrum and a coefficient of correlation COR(k,k) of amplitude of each line k with the homologous lines of the transmission spectra contained in the determined frequency band.

8. A device according to claim 7, wherein the extracting device comprises a block of filters for the determining of primary transmission parameters coupled with an intermediate computation block and a transmission detection block.

9. A device according to claim 8, wherein the block of filters, the intermediate computation block and the transmission detection block are formed by signal processors.

10. A device according to claim 9, wherein the signal processors are programmed to compute, for each transmission spectrum line k, a mean amplitude of the order k lines of the spectra contained in the frequency band, a signal-to-noise ratio RSBk, a standard deviation ETk in amplitude and a coefficient of correlation COR(k,k) of amplitude with the homologous lines of the transmission spectra contained in the frequency band.