GB2512093A

GB2512093A - Scheme for detection and classification of modulated swept frequency signals

Info

Publication number: GB2512093A
Application number: GB1305086.9A
Authority: GB
Inventors: Keith John Jones
Original assignee: TRL Technology Ltd
Current assignee: L3Harris TRL Technology Ltd
Priority date: 2013-03-20
Filing date: 2013-03-20
Publication date: 2014-09-24
Anticipated expiration: 2033-03-20
Also published as: GB201305086D0; GB2512093B

Abstract

Individual signals are extracted from received radio frequency signals and analysed to confirm whether any are of the modulated swept frequency type. The analysis may comprise analysing the spectral size and shape of the individual extracted signals and computing the autocorrelation function of the extracted signal data to determine modulation-independent parameters of the extracted individual signals. When the signal is of the modulated swept frequency type, the relevant modulation parameters (such as phase and frequency offsets of each subsequent frequency sweep of the extracted individual signals relative to initial phase and frequency) defining the signal are extracted so that a parameter set might be compared with predefined operating characteristics that correspond to modulated swept frequency signals of known origin. As a result, the modulated swept frequency signal generation device from which the received signals emanate may be subsequently classified as being either known or unknown type, which may be used by an unfriendly third party for the purposes of jamming communication signals.

Description

Scheme for Detection and Classification of Modulated Swept Frequency Signals This invention relates to signal processing methods and apparatus.

Ba ckp round In various applications it is useful to have a good knowledge, in terms of both origin (i.e. whether the associated signal generation device is known' or unknown') and spectral location (i.e. centre frequency and bandwidth), of each signal present within the signal environment. This includes both the radar and sonar fields, where a standard unmodulated swept frequency signal -which may be simply regarded as a special case of the modulated version of the swept frequency signal (i.e. where the modulation parameters are zero-valued) and is often referred to in the technical literature as a chirp' -may be used by a third party for the purposes of pulse compression, as well as that of wireless communications, where a modulated version of the swept frequency signal might typically be used by an unfriendly third party for the purposes of jamming one's own communication signals. A typical frequency range of interest for a radio frequency application might be from a few MHz up to several GHz.

In more detail, a problem exists in the fields of signal surveillance and intelligence whereby it is required to detect each signal present within the signal environment and to determine whether the signal generation device from which a particular signal emanates is of known or unknown origin. The modulated swept frequency signal is an important one in many application areas, as discussed above, so that there is a need to be able to detect such a signal in a blind fashion, i.e. without any a priori' knowledge relating to the signal, and to be able to extract the relevant parameters defining the signal. If this could be achieved, the resulting modulation parameter set could then be used to determine if the signal derives from a known origin.

For example, in the realm of electronic warfare (EW), where one would expect the increasingly busy signal environment to be populated by a multiplicity of such jamming devices, there would be some merit in being able to place a listening device or receiver within that environment which would be able to capture all those signals of modulated swept frequency type. One would then be able to measure and exploit various signal parameters, including the initial phases and/or frequencies of the multiple frequency sweeps, so as to determine which, of the multiplicity of jamming devices present within the signal environment, might be considered to be of known origin and which, conversely, might be considered to be of unknown origin.

It is therefore an aim of the present invention to provide methodology and apparatus to effect such parameter extraction, and further to enable such identification to determine whether signals present within the signal environment are from known or unknown origins. A typical frequency range of interest for a radio frequency application might be from a few MHz up to several GHz, and the term "radio frequency" as used herein is understood to encompass signals within that range.

In accordance with the present invention, this aim is achieved by a multi-stage process of extracting any individual signal from a received wideband signal that resembles' a signal of modulated swept frequency type, determining the modulation-independent parameters of those received signals to assess whether those signals are indeed of the modulated swept frequency type, and then, when appropriate, extracting the modulation-dependent parameters. Those parameters may then be compared with pre-defined operating characteristics to determine if the signals originate from known or unknown sources.

In accordance with a first aspect of the present invention there is provided a signal processing method for identifying modulated swept frequency signals from a signal environment, comprising the steps of: a) receiving radio frequency signals from the signal environment across a frequency band using a radio frequency signal receiving means; b) extracting individual signals present from within the received signals; c) analysing each of the extracted individual signals to confirm if any of those extracted individual signals are of the modulated swept frequency type; and d) determining the modulation-dependent parameters for each extracted individual signal which is confirmed as being of the modulated swept frequency type.

In accordance with a second aspect of the present invention there is provided a method for determining whether the source of a data signal is known or unknown, comprising the signal processing method of the first aspect, and further comprising the step of: e) comparing the modulation-dependent signal parameters determined in step d) to known values.

In accordance with a third aspect of the present invention there is provided apparatus for performing the signal processing method in accordance with either of the first or second aspect.

In accordance with a fourth aspect of the present invention there is provided apparatus for performing signal processing for identifying modulated swept frequency signals from a signal environment, comprising: Is receiving means for receiving radio frequency signals from the signal environment across a frequency band; first processing means for extracting individual signals present from within the received signals; second processing means for analysing each of the extracted individual signals to confirm if any of those extracted individual signals are of the modulated swept frequency type; and third processing means for determining the modulation-dependent parameters for each extracted individual signal which is confirmed as being of the modulated swept frequency type.

According to a preferred embodiment, the present invention may provide: 1. A processing system that is able firstly to extract from a wideband signal received by one's own electronic equipment any individual signal present within the signal environment that resembles a signal of modulated swept frequency type. Upon completion of this stage, the processing system then determines the modulation-independent parameters of the signal so as to assess whether or not the signal is of modulated swept frequency type. Finally, if the signal is of modulated swept frequency type, the processing system then extracts the modulation-dependent parameters of the signal, thus enabling the origin of the signal (i.e. the associated signal generation device) to be subsequently classified as being of either known or unknown type.

2. A processing system, as set out above, that takes the multiple data streams produced by a suitably defined channelizer from the signal produced by the analogue-to-digital conversion (ADC) unit of one's own receiver and extracts from the multiple data streams those signals that resemble' signals of modulated swept frequency type.

3. A processing system, as set out above, that computes the autocorrelation function (ACF) of the extracted signal data and from the ACF determines the modulation-independent parameters of the signal -namely the period of the modulated swept frequency signal and the number of frequency sweeps -when appropriate, given that the signal may or may not turn out to be of modulated swept frequency type, so that one may assess whether or not the signal actually is of modulated swept frequency type.

is 4. A processing system, as set out above, that determines the modulation-dependent parameters of the modulated swept frequency signal -namely the phase and frequency offsets -so that the origin of the signal (i.e. the associated signal generation device) might be subsequently classified as being of either known or unknown type.

5. A processing system, as set out above, that: 1) segments and overlaps the sampled data stream where the segment length is dependent upon the stationarity of the anticipated signals and the amount of overlap is dependent upon the anticipated level of signal corruption at the boundary regions due to the familiar problem of end effects'; 2) performs the fast Fourier transform (FFT) upon each data segment so as to convert the signal from the time domain to the frequency domain and computes estimates for both the lower' and upper' boundaries of each signal spectrum from analysis of the resulting power spectral density (PSD) measurements; 3) produces a separate version of the spectrum for each signal whose spectral characteristics -namely size and shape -suggests that it might be of modulated swept frequency type and filter off the unwanted spectral components through the modification of the spectral samples; 4) performs an inverse FFT, or IFFT, for each version of the spectrum so as to convert the signal from the frequency domain to the time domain and discards the boundary regions of the resulting time-domain signal segment in order to address the familiar problem of end effects' -the overall size of the discarded boundary regions should be equal to the amount of overlap used in the initial segmentation process; and 5) computes estimates for both the lower' and upper' boundaries of each signal from analysis of the resulting signal magnitude measurements made in the time domain and extract the signals for subsequent analysis.

6. A processing system, as set out above, that: 1) computes the ACF from the captured signal segment, then, given that the signal in question may -at least potentially at this stage -be of modulated swept frequency type: 2) computes the best' estimate for the period of the modulated swept frequency signal, as measured in samples, using the number-theoretic properties of the periodicities of the ACE peaks; and 3) using the estimates already obtained for the boundaries and period of the modulated swept frequency signal computes the corresponding best' estimate for the number of frequency sweeps generated.

7. A processing system, as set out above, that: 1) using the estimates for the signal boundaries, the swept frequency period and the number of frequency sweeps generated, computes the relative phase offset and the relative frequency offset for each odd-numbered frequency sweep relative to those for the first frequency sweep and for each even-numbered frequency sweep relative to those for the second frequency sweep; and 2) compares the resulting sets of relative phase offsets and relative frequency offsets against those stored in the look-up tables in order to determine whether the swept frequency signal generation device from which the signal emanates is of known or unknown type.

For the final comparison step, where signal parameters are compared with pre-defined operating characteristics, these characteristics may be stored by the user, e.g. in the form of look-up tables, that correspond to all those modulated swept frequency signals present within the signal environment that are of known origin. If there is concordance between the extracted signal parameters and stored characteristics then the signal generation device from which the received signal emanates would be classified as being of known type, while the absence of such concordance would lead to the signal generation device being classified as being of unknown type.

Such a system enables one to firstly determine, without any a priori' knowledge relating to the signal, whether a received signal is of modulated swept frequency type. When the signal is of modulated swept frequency type, the scheme further enables one to extract the relevant modulation parameters defining the signal so that the parameter set might be compared with those pre-defined operating characteristics, e.g. as stored in the form of look-up tables, which correspond to modulated swept frequency signals of known origin. As a result, the modulated swept frequency signal generation device from which the received signal emanates may be subsequently classified as being of either known or unknown type.

Theory Before turning to the present invention proper, some background theory is hereby presented for clarity.

Note that since the standard unmodulated swept frequency signal is simply a special case of the modulated swept frequency signal, where the phase and/or frequency modulation is trivial, i.e. the modulation parameters are zero-valued, we will hereafter use the term modulated swept frequency signal' to represent any signal that is of swept frequency type, regardless of whether the modulation is trivial or non-trivial.

For a brief mathematical description of the standard unmodulated swept frequency signal, it may be written in its most generic form as y(f f. , t) = A(t).e(2ttk0) . . (1) where A(t)' is the time-varying amplitude -generally regarded as piecewise constant (at least approximately) for most applications of interest, the carrier frequency, 4' the initial phase and P(f11.t)' the time-varying phase function, as derived from the time-varying modulation frequency via the expression (t)Umt) 27t!m(t).th.

As a result, for the particular case where the frequency varies linearly with respect to time, whereby Id fill (t) = -i---j- t) = with the rate of change of frequency, so the phase function may be written as (m' t) = 2irf3t2 /2 = itf3t2. . . . (4) via the integral expression of Equation 2.

In the following text, reference will be made to: [1] F.J. Harris, "Multirate Signal Processing for Communication Systems", Prentice Hall, New Jersey, 2004.

[2] A.V. Oppenheim & R.W. Schafer, "Discrete-Time Signal Processing", Prentice Hall, New Jersey, 1989.

Detailed descrirjtion The invention will now be described with reference to the accompanying drawings, in which: Fig. 1 schematically shows a processing system in accordance with the present invention; Fig. 2 schematically shows a scheme for PDFT channelization of a received signal as employed by the present invention; Fig. 3 schematically shows a signal extraction routine as employed by the present invention; Fig. 4 schematically shows a modulation-independent parameter estimation routine as employed by the present invention; Fig. 5 schematically shows a modulation-dependent parameter estimation routine as employed by the present invention; and Fig. 6 schematically shows the interaction of individual components of a device in accordance with the present invention.

A processing system in accordance with the present invention is schematically shown in Fig. 1, which illustrates how the five main functions to be performed by the present invention (namely the wideband signal channelization, the signal extraction, the modulation-independent parameter estimation, the modulation-dependent parameter estimation and the signal origin determination (for those signals found to be of modulated swept frequency type)) interact with each other and with the outside world. The main components required to perform these tasks are described further below, with reference to Fig. 6.

Initially, it is essential to receive wideband radio frequency signals from the signal environment, using a radio frequency receiver. Analogue-to-digital conversion of the Is received signals, using an analogue-to-digital converter, produces a sampled data stream, and this input is shown as "from wideband ADC" in Fig. 1.

A first main phase of the inventive system may be considered to be one of capturing those individual signals present within the signal environment that resemble signals of modulated swept frequency type. This is achieved by partitioning the received wideband signals such that a unique time series is obtained for each individual signal whose spectral characteristics -namely size and shape -suggests that it might be of modulated swept frequency type. Components are required to carry out the following functions: -Perform channelization upon the sampled data stream produced from the analogue-to-digital conversion (ADC) unit of a receiver using a suitably defined channelizer' -this being the term commonly used in the technical literature to denote a device that performs the act of channelization, namely that partitions a signal over frequency such that a unique signal is produced for each frequency interval where that frequency interval is in turn referred to as a channel'. Typical channelizers are obtained with a bank of digital down conversion (DDC) units or, for the case of a wideband received signal, with a computationally efficient polyphase discrete Fourier transform (DFT) or PDFT [1]. A PDFT is shown in Fig. 2, where a band-pass complex-valued input signal {x[n]} is split, down-sampled and filtered before undergoing N-point complex-data discrete Fourier transformation to produce a set of low-pass complex-valued output signals.

Then, for each channelized data stream:- -Segment and overlap the sampled data stream where the segment length is dependent upon the stationarity [2] of the anticipated signals and the amount of overlap is dependent upon the anticipated level of signal corruption at the boundary regions due to the familiar problem of end effects' [2]. The end effects problem is often referred to in the technical literature as the Gibbs phenomenon and refers to the oscillatory behaviour occurring near the discontinuous boundaries of the signal segment being processed. The signal corruption is further increased here by the need for subsequent spectral filtering -to be discussed later -so that a large overlap of the order of 75% may well be required to ensure that possible artefacts are removed or minimized prior to subsequent signal reconstruction. Note that the duration of the segment length must not be greater than the stationarity of the signal -which as is known in the art is the time period over which the statistics of the signal, such as those of mean and variance, would be expected to remain approximately constant -otherwise the performance of the subsequent signal reconstruction will be degraded.

Perform a fast Fourier transform (FF1) upon each data segment so as to convert the signal from the time domain to the frequency domain and compute estimates for both the lower' and upper' boundaries of each signal spectrum from analysis of the resulting power spectral density (PSD) measurements. This indicates the spectral width -whose measurement will be discussed later-of the signal.

Produce a separate version of the spectrum for each signal whose spectral characteristics -namely size and shape -suggests that it might be of modulated swept frequency type and filter off the unwanted spectral components through the modification of the spectral samples. With regard to the shape and size, the signal is regarded as having the required spectral characteristics if it comprises one spectral region whose width, in terms of spectral samples, is greater than some pre-specified value -a corresponding spectral bandwidth of 100 kHz, for example, might typically be used as a lower threshold. The spectral filtering is achieved by applying a spectral window with tapered edges to the signal spectrum, this minimizing spectral discontinuities and ensuring that artefacts occurring in the subsequent signal reconstruction are kept to a minimum. The filtering off of the unwanted regions of the spectrum also has the desirable effect of increasing the signal-to-noise ratio (SNR) after the subsequent signal reconstruction has taken place, thereby enhancing the likely performance of subsequent processing algorithms which will inevitably be dependent upon SNR.

-Perform an inverse EFT or lEFT, for each version of the spectrum, so as to convert the signal from the frequency domain to the time domain and discard the boundary regions of the resulting time-domain signal segment in order to address the familiar problem of end effects' -the overall size of the discarded boundary regions should be equal to the amount of overlap used in the initial segmentation process.

-Compute estimates for both the lower' and upper' boundaries of each signal from analysis of the resulting signal magnitude measurements made in the time domain and extract the signals for subsequent analysis. The lower' and upper' boundaries are thus obtained from an analysis of the signal's leading and trailing edges, respectively, with amplitude thresholds being used to determine where the signal starts and ends.

The above method thus enables each of one or more channelized data streams, i.e. one stream per channel, to be itself decomposed into multiple data streams where each data stream may correspond -at least potentially at this stage -to a unique modulated swept frequency signal and where multiple segments of data corresponding to the same modulated swept frequency signal may be subsequently pieced together in a relatively seamless fashion via application of an overlap-save' signal reconstruction technique, as known per se from [2]. As the resulting data streams are independent of each other, the subsequent processing tasks to be performed upon the multiple data streams may be carried out in a parallel fashion given the availability of suitable parallel processing equipment onto which those tasks might be mapped.

Fig. 3 details the signal extraction routine in more detail, showing, from the top, a segmented and overlapped channel time series, which is then subjected to an FF1.

This produces a channel spectrum comprising multiple swept frequency signals, Sn, multiple communication signals, C, and possibly some additional signals not of interest. Spectrum partitioning and replication produces separate spectra for each swept frequency signal, with those for S1 and 52 shown. Each of these spectra is subjected to an IFFT, to produce respective corrected time series for that signal.

Boundary regions are discarded as described above to produce non-overlapping time series.

The resulting reconstructed signal data is assumed to be complex valued, possessing both amplitude and phase information, in order that the subsequent boundary estimation and parameter estimation algorithms be effectively and straightforwardly applied. When the sampled data produced by the channelizer is complex valued, this is automatically the case. When the sampled data produced by the channelizer is real valued -as produced by the PDFT for the channel centred on zero frequency -possessing just amplitude information, the reconstructed signal data may be straightforwardly converted to complex-valued form through the zeroization of the negative frequency spectral components, following the application of the FFT, as this has the same effect upon the sampled data as that of the Hilbert transform [1] operation which is the most commonly used tool for converting real-valued signals to complex-valued form.

The second main phase of the invention may be considered to be one of determining the modulation-independent parameters of the signal -when appropriate, given that the signal may or may not turn out to be of modulated swept frequency type -so as to assess whether or not the signal is of modulated swept frequency type. A schematic routine showing this phase is shown in Fig. 4, which is capable of determining those signal parameters associated with a given modulated swept frequency signal that remain independent of any phase and/or frequency modulation present on the signal, namely those that relate to the period of the swept frequency signal and to the number of frequency sweeps generated. The system comprises components which are able to carry out the following functions: -Compute the autocorrelation function (ACF) from the captured signal segment.

Then, given that the signal in question may -at least potentially at this stage - be of modulated swept frequency type:- -Compute the best' estimate for the period of the modulated swept frequency signal, as measured in samples, using statistical analysis of the number-theoretic properties of the periodicities of the ACF peaks -this computation is important, and described in more detail below.

-Using the estimates already obtained for the boundaries and period of the modulated swept frequency signal compute the corresponding best' estimate for the number of frequency sweeps generated.

The above system thus produces estimates -for signals which are of modulated swept frequency type -for those signal parameters that are independent of any non-trivial modulation present on the swept frequency signal, but which will nevertheless be essential for the subsequent determination of those signal parameters that actually define the modulation.

With regard to the step of computing the best estimate of the signal period, in the absence of modulation on the swept frequency signal, the ACF peaks will be both strong (in terms of magnitude) and equally spaced (with spacing equal to the period of the swept frequency signal), whereas when modulation is present there will be increased variability -in terms of both strength and spacing -of the ACF peaks. To address this problem, it will be noted that for each pair of frequency sweeps, with two periods of separation, the phase differences between consecutive samples will occur at the same point within the frequency sweep, regardless of the sweep pattern used, i.e. regardless of how the sweep repeats. This property may therefore be used to determine a second estimate of the number samples per swept frequency period which, together with the estimate of the spectral width of the signal (already computed in the first phase), can be used to enhance the performance of the ACF-based technique and thus to increase the robustness of the modulation-dependent parameter estimation process that follows.

It should be noted that with regard to the ACE-based technique, the best estimate of the signal period is defined as being that which leads to the most stable' set of ACF peaks, i.e. where the peaks are sufficiently strong (exceeding some specified threshold relative to the main peak) and where the positions of the peaks (in terms of samples offset from the position of the main peak) most closely resemble those corresponding to the positions of the measured peaks. When the ACE peaks do not obey such a pattern the signal is assumed to be not of modulated swept frequency type and no further processing is performed upon the signal.

The third main phase of the invention may be considered to be one of determining the modulation-dependent parameters of the modulated swept frequency signal so that the origin of the signal (i.e. the associated signal generation device) might be subsequently classified as being of either known or unknown type. A system for enabling the determination of those signal parameters associated with a given modulated swept frequency signal that relate to the presence of any non-trivial phase and/or frequency modulation, namely the initial phases and/or frequencies associated with the multiple frequency sweeps of the signal, is schematically shown in Fig. 5. The system makes use of components which are able to carry out the following functions: -Using the estimates for the lower' and upper' signal boundaries, as obtained from the reconstructed time-domain signal, the swept frequency period and the number of frequency sweeps generated -as discussed earlier in the detailed description -compute the relative phase offset and the relative frequency offset for each odd-numbered frequency sweep relative to those for the first frequency sweep and for each even-numbered frequency sweep relative to those for the second frequency sweep. This is another important step of the invention and is described in considerably more detail below; and, -subsequently compare the resulting sets of relative phase offsets and relative frequency offsets against those stored, for example in look-up tables or similar database, in order to determine whether the swept frequency signal generation device from which the signal emanates is of known or unknown type.

The above system thus produces estimates for those signal parameters that define the modulation on the swept frequency signal and which thus enable the origin (i.e. whether the signal generation device is known or unknown) of the modulated swept frequency signal to be determined. With regard to this third phase, a single procedure is defined which is able to deal with all three possible modulation combinations: 1) that arising from the modification of the initial phase of each frequency sweep; 2) that arising from the modification of the initial frequency of each frequency sweep; and 3) that arising from the simultaneous modification of both the initial phase and the initial frequency of each frequency sweep.

To achieve this, an assumption is first made that any modulated swept frequency signal will possess modifications to both the initial phases and the initial frequencies but that either one or both sets of modifications may correspond to the trivial case of zero-valued modifications; clearly, when both sets of modifications are zero-valued, the signal reduces to the standard unmodulated swept frequency signal. Also, because of the time-invariant nature of the system represented mathematically by the complex exponential function that has been used to generate the modulated swept frequency signal (see Equation 1) modifications to the initial phase, when applied, will propagate through the frequency sweep so that every signal sample within the frequency sweep will undergo the same change in phase regardless of how the frequency of the signal varies with respect to time.

Thus, in accordance with the present invention, a three-stage processing system has been described that is able to: 1) capture all those individual signals present within the signal environment that resemble signals of modulated swept frequency type; and, for each such signal, to: 2) determine the modulation-independent parameters of the signal (when appropriate, given that the signal may or may not turn out to be of modulated swept frequency type) so as to assess whether or not the signal is of modulated swept frequency type; and, for each signal that is of modulated swept frequency type, to 3) determine the modulation-dependent parameters of the signal.

On completion of these three stages, the resulting parameter set for each captured signal of modulated swept frequency type may then be compared with pre-defined operating characteristics, as typically stored by the user in the form of look-up tables, so as to determine whether the signal generation device from which the received signal emanates may be subsequently classified as being of either known or unknown type.

Further Discussion As mentioned above, further detail is now given as to how, in the third phase above, the estimates for the signal boundaries, the swept frequency period and the number of frequency sweeps generated may be used to compute the relative phase offset and the relative frequency offset for each odd-numbered frequency sweep relative to those for the first frequency sweep and for each even-numbered frequency sweep relative to those for the second frequency sweep.

To see the truth of the time-invariance property of the system represented mathematically by the complex exponential function of Equation 1, suppose that a phase shift, D'is applied to the first sample of the frequency sweep so that the initial phase becomes 4 + 4' instead of. Then expanding out the resulting expression produces so that delaying the input to the system (in time or, equivalently, phase) results in a similar delay in the output from the system at any given time (or, equivalently, frequency), i.e. the same phase shift appears on every sample of the frequency sweep, identical to that applied to the first, as required.

Before the procedure starts, pre-defined operating characteristics need to be stored in look-up tables so that: a) the range of possible relative initial phases is represented by means of a set of values which may be read from a look-up table and which when each phase value is applied as a phase rotation to the respective frequency sweeps of the swept frequency signal reduce the initial phases of each frequency sweep by commensurate amounts, whilst: b) the range of possible relative initial frequencies is represented by means of a set of values for which a corresponding set of complex exponential demodulating sequences have been created which may be read from a look-up table and which when each demodulating sequence is applied, sample by sample, to the respective frequency sweeps of the modulated swept frequency signal reduce the initial frequencies of each frequency sweep by commensurate amounts without altering the value of the initial phase.

For a brief description of a typical demodulating sequence and how it works, if a modulated swept frequency signal needs to be down shifted for a given frequency sweep by a frequency 4Tfl" say, without altering the value of the initial phase, then the sequence {clrn[n]} I. . (6) if applied sample-by-sample to the modulated swept frequency signal for that particular frequency sweep, will achieve this, where öt' is the sampling interval and with n varying from 0 to N-i with N' the number of samples per frequency sweep.

Multiple demodulating sequences, such as that given by Equation 6 above, therefore need to be replicated for a number of possible frequencies in order that the required frequency range is adequately covered.

It should be noted that the look-up tables need to store relative initial phases and relative initial frequencies, rather than initial phases and initial frequencies, as the measured values of the actual signal phase will be dependent upon the elapsed time or delay between the signal being transmitted and the signal being received and therefore of no use for classification purposes. The adoption of relative rather than actual values overcomes this problem as they are independent of the time at which the signal was transmitted and/or received.

At this point, it is necessary to distinguish between two possible modulated swept frequency signal types: 1) where for every frequency sweep the frequency of the signal rises from the lowest frequency in its frequency range to the highest frequency in its frequency range -referred to as an "up-sweep", or, where for every frequency sweep the frequency of the signal falls from the highest frequency in its frequency range to the lowest frequency in its frequency range -referred to as a "down-sweep"; and 2) where for every odd-numbered frequency sweep the frequency of the signal rises from the lowest frequency in its frequency range to the highest frequency in its frequency range (i.e. an up-sweep), whilst for every even-numbered frequency sweep the frequency of the signal falls from the highest frequency in its frequency range to the lowest frequency in its frequency range (i.e. a down-sweep); or vice versa.

From the above definitions of the two possible modulated swept frequency signal types, it may be inferred that in the absence of modulation and noise, the signal segments for all the odd-number frequency sweeps will be identical and the signal segments for all the even-numbered frequency sweeps will be identical, regardless of which swept frequency signal type is used, i.e. whether the up-up' or down-down' of type 1) or the up-down' or down-up' of type 2), where up-up' corresponds to the case where all frequency sweeps are rising, down-down' to the case where all frequency sweeps are falling, and up-down' or down-up' to the case where the frequency sweeps alternate between rising and falling. As a result, by comparing on a sample-by-sample basis the phase values of all the odd-numbered frequency sweeps (apart from the first) with those of the first frequency sweep and the phase values of all the even-numbered frequency sweeps (apart from the second) with those of the second frequency sweep, it is possible to determine what modifications, if any, might have been made to the initial phases of each frequency sweep, in Is relative terms, when compared to the first two frequency sweeps. Before such a test can be carried out, however, it is first necessary to be able to address the problem of possible modifications to the initial frequencies as the start frequencies of all the odd-numbered frequency sweeps will need to be frequency aligned before any phase comparison can be carried out, with the same being true for the case of all the even-numbered frequency sweeps.

Thus, to determine the best estimates for the relative initial phases and relative initial frequencies the algorithm seeks to test the measurements against all possible combinations of values from the look-up tables, relative initial frequency followed by relative initial phase, keeping track of the best' match as it does so. Thus, for a given frequency sweep, the algorithm first selects a possible relative frequency offset, applying the corresponding demodulating sequence to the samples of the frequency sweep, before iteratively applying phase rotations to account for all possible relative phase offsets, keeping track of the best match as it does so. Once the algorithm has exhausted all possible combinations of relative frequency offset and relative phase offset the best' solution is obtained, this corresponding to the case where the averaged relative phase error over each frequency sweep is minimized. This procedure is repeated for each set of modulation-independent parameters determined during the second phase and the resulting solution sets compared against those stored in the look-up tables to determine whether the modulated swept frequency signal generation device from which the signal emanates is of known or unknown type.

To simplify the comparisons with the contents of the look-up tables, the relative phase offsets and relative frequency offsets for each frequency sweep may be straightforwardly converted from reference to the initial frequency sweep (the first frequency sweep for the odd-numbered frequency sweeps or the second frequency sweep for the even-numbered frequency sweeps) to reference to its predecessor so that one look-up table needs to contain two sets of interleaved phase differences, one set for the odd-numbered frequency sweeps and one set for the even-numbered frequency sweeps, whilst the other look-up table needs to contain two sets of interleaved frequency differences, one set for the odd-numbered frequency sweeps and one set for the even-numbered frequency sweeps.

With regard to this aspect, two possible sources of error may compromise the performance of the algorithm in its ability to correctly classify the modulated swept frequency signal:-the first relates to the possible error in the estimation of the lower' and upper' boundaries of the signal, since if one or other of these is sufficiently inaccurate the algorithm will look to compare sections of signal' against sections of non-signal'; and the second relates to the possible presence within a multi-path signal environment of non-direct path signal components which may lead to the algorithm looking to compare a single direct path signal component' against that of combined direct path + non-direct path signal components'. To address these two possible sources of error, adjustable safety margins' are defined at the beginning and end of each frequency sweep. The phase comparisons are then carried out for each frequency sweep using all samples other than those to be found in the safety margins, where the size of the safety margins is set by the user to cater for the pre-determined accuracy of the signal boundary estimation process as well as the maximum anticipated multi-path delay (as obtained from environmental intelligence) in samples. As a result, the algorithm is made robust in the sense that its performance becomes reasonably insensitive to the presence of such errors, provided that a sufficient number of samples are left between the two safety margins to enable statistically meaningful results to be obtained from the averaging of the relative phase errors.

Whole system Fig. 6 schematically shows the main components that make up an exemplary device in accordance with the present invention for performing the above phases, and their interactions. A control processor is linked to both FFT and PDFT processors, to an ACF processor and to a combined phase-frequency offset estimation (CPFOE) processor, as well as to a global memory which contains that data needing to be accessed by all the processors. The PDFT processor, which operates continuously on the received wideband signal, performs the task of wideband signal channelization, whereby the received signal is partitioned into a number of independent frequency bands which may be subsequently analyzed, independently of each other, for the presence of signals of interest. The FFT processor, as well as being utilised as a processing module by the FDFT processor, first assists the control processor in performing the task of extracting all those signals present within the signal environment that resemble signals of modulated swept frequency type. The ACF processor then performs the task of computing the ACF on the extracted signal data, with which the control processor then determines the modulation-independent parameters of the signal, when appropriate, given that the signal may or may not turn out to be of modulated swept frequency type -at this point we are able to assess whether or not the signal is of modulated swept frequency type. Finally, if the signal is of modulated swept frequency type, the CPFOE processor then determines the modulation-dependent parameters of the signal which, if non-zero, enables the control processor to determine the origin (i.e. whether the signal generation device is known or unknown) of the signal. It should be noted that each of the above processors preferably has access to its own local memory in order to maximize the speed with which the associated computations are performed. The control processor, as well as performing those individual tasks already defined, controls the flow of data between the other processors and between itself and the outside world and ensures that the operation of the other processors is appropriately synchronised so as to deliver the correct outputs.

The above-described embodiments are exemplary only, and other possibilities and alternatives within the scope of the invention will be apparent to those skilled in the art. For example, the exemplary embodiment described above uses a two-stage process to identify signals which are of the modulated swept frequency type, a first "coarse" assessment performed by looking at the spectral characteristics of transformed signals, and a second identification stage which involves determining modulation-independent parameters of the signals. While this is considered to be most computationally efficient, the initial coarse assessment may be omitted, in which case the determination of modulation-independent parameters would be carried out for the whole range of signals occurring within the received wideband signal. Clearly this would require greater signal processing power, but could be used, particularly if the wideband signal of interest is relatively narrow.

Claims

Claims 1. A signal processing method for identifying modulated swept frequency signals from a signal environment, comprising the steps of: a) receiving radio frequency signals from the signal environment across a frequency band using a radio frequency signal receiving means; b) extracting individual signals present from within the received signals; c) analysing each of the extracted individual signals to confirm if any of those extracted individual signals are of the modulated swept frequency type; and d) determining the modulation-dependent parameters for each extracted individual signal which is confirmed as being of the modulated swept frequency type.
2. A method according to 1, wherein step a) further comprises performing analogue-to-digital conversion of the received signals, using an analogue-to-digital converter, to produce a sampled data stream.
3. A method according to claim 2, wherein step b) comprises channelization of the data stream to produce multiple data streams.
4. A method according to claim 3, wherein the channelization is performed by applying a polyphase discrete Fourier transform to the data stream.
5. A method according to any preceding claim, wherein step c) comprises the step of: ci) initially analysing each of the extracted individual signals to identify any of those extracted individual signals indicative of being of the modulated swept frequency type.
6. A method according to claim 5, wherein step ci) comprises analysing spectral characteristics of the extracted individual signals.
7. A method according to claim 6, wherein the spectral characteristics analysed comprise the spectral size and shape of the extracted individual signals.
8. A method according to either of claims 6 and 7, wherein the spectral characteristics analysed comprise the spectral width of the extracted individual signals.
9. A method according to any preceding claim, wherein step c) comprises the step of: c2) determining modulation-independent parameters of each of the extracted individual signals to confirm if the extracted individual signal is of modulated swept frequency type.
10. A method according to claim 9 when dependent on claim 5, wherein step c2) comprises determining modulation-independent parameters only of the signals identified in step ci) to confirm if the signal is of modulated swept frequency type.i
ii. A method according to either of claims 9 and 10, wherein step c2) comprises computing the autocorrelation function of the extracted signal data and from the autocorrelation function determines the modulation-independent parameters of the extracted individual signal.12. A method according to any preceding claim, wherein the modulation-dependent parameters determined in step d) comprise the initial phase associated with the multiple frequency sweeps of the extracted individual signal.13. A method according to claim 12, wherein step d) further comprises determining the phase offset of each subsequent frequency sweep of the extracted individual signal relative to the initial phase.14. A method according to any preceding claim, wherein the modulation-dependent parameters determined in step d) comprise the initial frequency associated with the multiple frequency sweeps of the extracted individual signal.15. A method according to claim 14, wherein step d) further comprises determining the frequency offset of each subsequent frequency sweep of the extracted individual signal relative to the initial frequency.16. A method for determining whether the source of a data signal is known or unknown, comprising the signal processing method of any preceding claim, and further comprising the step of: e) comparing the modulation-dependent signal parameters determined in step d) to known values.17. A method according to claim 16, wherein the known values are stored in alook-up table.18. A method according to any preceding claim, wherein step b) is performed using a polyphase discrete Fourier transform processor.19. A method according to any of claims 5 to 18, wherein step ci) is performed using a fast Fourier transform processor.20. A method according to any of claims 5 to 19, wherein step c2) is performed using an autocorrelation function processor.21. A method according to any preceding claim, wherein step d) is performed using a combined phase frequency offset estimation processor.22. Apparatus for performing the signal processing method in accordance with any preceding claim.23. Apparatus for performing signal processing for identifying modulated swept frequency signals from a signal environment, comprising: receiving means for receiving radio frequency signals from the signal environment across a frequency band; first processing means for extracting individual signals present from within the received signals; second processing means for analysing each of the extracted individual signals to confirm if any of those extracted individual signals are of the modulated swept frequency type; and third processing means for determining the modulation-dependent parameters for each extracted individual signal which is confirmed as being of the modulated swept frequency type.24. Apparatus according to claim 23, wherein the first processing means comprises a polyphase discrete Fourier transform processor.25. Apparatus according to either of claims 23 and 24, wherein the second processing means comprises a fast Fourier transform processor and an autocorrelation function processor.26. Apparatus according to any of claims 23 to 25, wherein the third processing means comprises a combined phase frequency offset estimation processor.27. A method substantially as herein described with reference to the accompanying figures.28. Apparatus substantially as herein described with reference to the accompanying figures.Amendments to the Claims have been filed as follows Claims 1. A signal processing method for identifying modulated swept frequency signals from a signal environment, comprising the steps of: a) receiving radio frequency signals from the signal environment across a frequency band using a radio frequency signal receiving means; b) extracting individual signals present from within the received signals; c) analysing each of the extracted individual signals to confirm if any of those extracted individual signals are of the modulated swept frequency type; and d) determining the modulation-dependent parameters for each extracted individual signal which is confirmed as being of the modulated swept frequency type; wherein the modulation-dependent parameters determined in step d) comprise the initial phase associated with the multiple frequency sweeps of the extracted individual signal, and / or the initial frequency associated with the multiple frequency sweeps of the extracted individual signal.o 2. A method according to 1, wherein step a) further comprises performing analogue-to-digital conversion of the received signals, using an analogue-to-digital r-20 converter, to produce a sampled data stream.3. A method according to claim 2, wherein step b) comprises channelization of the data stream to produce multiple data streams.4. A method according to claim 3, wherein the channelization is performed by applying a polyphase discrete Fourier transform to the data stream.5. A method according to any preceding claim, wherein step c) comprises the step of: ci) initially analysing each of the extracted individual signals to identify any of those extracted individual signals indicative of being of the modulated swept frequency type.6. A method according to claim 5, wherein step ci) comprises analysing spectral characteristics of the extracted individual signals.7. A method according to claim 6, wherein the spectral characteristics analysed comprise the spectral size and shape of the extracted individual signals.8. A method according to either of claims 6 and 7, wherein the spectral characteristics analysed comprise the spectral width of the extracted individual signals.9. A method according to any preceding claim, wherein step c) comprises the step of: c2) determining modulation-independent parameters of each of the extracted individual signals to confirm if the extracted individual signal is of modulated swept frequency type.ct 10. A method according to claim 9 when dependent on claim 5, wherein step c2) o comprises determining modulation-independent parameters only of the signals identified in step ci) to confirm if the signal is of modulated swept frequency type.

ii. A method according to either of claims 9 and 10, wherein step c2) comprises computing the autocorrelation function of the extracted signal data and from the autocorrelation function determines the modulation-independent parameters of the extracted individual signal.
12. A method according to any preceding claim, wherein the modulation-dependent parameters determined in step d) comprise the initial phase associated with the multiple frequency sweeps of the extracted individual signal.
13. A method according to claim 12, wherein step d) further comprises determining the phase offset of each subsequent frequency sweep of the extracted individual signal relative to the initial phase.
14. A method according to any preceding claim, wherein the modulation-dependent parameters determined in step d) comprise the initial frequency associated with the multiple frequency sweeps of the extracted individual signal.
15. A method according to claim 14, wherein step d) further comprises determining the frequency offset of each subsequent frequency sweep of the extracted individual signal relative to the initial frequency.
16. A method for determining whether the source of a data signal is known or unknown, comprising the signal processing method of any preceding claim, and further comprising the step of: e) comparing the modulation-dependent signal parameters determined in step d) to known values.
17. A method according to claim 16, wherein the known values are stored in alook-up table.o
18. A method according to any preceding claim, wherein step b) is performed using a polyphase discrete Fourier transform processor.
19. A method according to any of claims 5 to 18, wherein step ci) is performed using a fast Fourier transform processor.
20. A method according to any of claims 5 to 19, wherein step c2) is performed using an autocorrelation function processor.
21. A method according to any preceding claim, wherein step d) is performed using a combined phase frequency offset estimation processor.
22. Apparatus for performing the signal processing method in accordance with any preceding claim.
23. Apparatus for performing signal processing for identifying modulated swept frequency signals from a signal environment, comprising: receiving means for receiving radio frequency signals from the signal environment across a frequency band; first processing means for extracting individual signals present from within the received signals; second processing means for analysing each of the extracted individual signals to confirm if any of those extracted individual signals are of the modulated swept frequency type; and third processing means for determining the modulation-dependent parameters for each extracted individual signal which is confirmed as being of the modulated swept frequency type, wherein the modulation-dependent parameters comprise the initial phase associated with the multiple frequency sweeps of the extracted individual signal and / or the initial frequency associated with the multiple frequency sweeps of the extracted individual signal.
24. Apparatus according to claim 23, wherein the first processing means comprises a polyphase discrete Fourier transform processor.o
25. Apparatus according to either of claims 23 and 24, wherein the second processing means comprises a fast Fourier transform processor and an -2O autocorrelation function processor.
26. Apparatus according to any of claims 23 to 25, wherein the third processing means comprises a combined phase frequency offset estimation processor.
27. A method substantially as herein described with reference to the accompanying figures.
28. Apparatus substantially as herein described with reference to the accompanying figures.