Mitigating interference experienced by a global navigation satelite system

Abstract

A device 10 for mitigating interference experienced by a receiver of a global navigation satellite system (GNSS) incorporates a GNSS antenna 12 whose received signals are down converted and then overs

Description

Mitigating Interference experienced by a Global Navigation Satellite System This invention relates to an apparatus and a method for mitigating interference experienced by a receiver of a global navigation satellite system. Global navigation satellite systems include the Global Positioning System (GPS), Galileo, GLONASS and Compass. GPS in particular consists of a set of satellites that transmit radio frequency (RF) signals containing information from which a GPS receiver can determine its position. These RF signals are very low in strength and signal to noise ratio (SNR) when received on the surface of the earth. Consequently, interference signals can easily have sufficient strength to prevent a receiver from detecting GPS signals and to deny it the ability to derive a position fix. A GPS satellite transmits RF signals on two carrier frequencies, L; at 1575.42 MHz and L, at 1227.6 MHz. Such a satellite has ephemeris and health data comprising a 1500 bit, 50 bps message modulated by two spreading codes. A 1.023 MHz clear / acquisition (C / A) code and a 10.23 MHz precision (P) code are used for the L, link while the P-code alone is used for the L, link. The P-code is encrypted, denoted as P(Y). The C / A code has a chip rate of 1.023 Mcps and a period of 1ms while the encrypted P(Y) has a chip period of 10.23 Mcps and a period of 1 week. GPS employs direct-sequence spread-spectrum (DS-SS) transmissions which allow pseudo-range measurements to a constellation of satellites to be made, which, together with the satellite ephemerides, allow an earth- based GPS receiver to establish its position. Despite the fact that GPS is a DS-SS system with a relatively large processing gain (typically 46 dB or so for the C / A code), its Immunity from interference is rather poor. This arises because much of the processing gain is required to raise the GPS signal above the receiver's thermal noise floor so that residual processing gain is small, Several techniques have been suggested in the scientific literature to enable interference to be removed and allow a GPS receiver to obtain a position fix. If an Inertial Navigation System (INS) is used with a GPS receiver, INS data allows the GPS receiver to use narrower band pass filters than Is usual: this allows a larger proportion of an interference signal to be rejected compared to that rejected by a conventional GPS receiver. RCIA ape Polarisation diversity processing may be used with a GPS antenna able to discriminate between right-hand and left-hand circularly polarised RF signals: something along these lines appears to be proposed by Braasch M.S. and Snyder C.A., ‘Running Interference: Testing a Suppression Unit, GPS World, Mar 1998, pp50-54). GPS signals are transmitted with right-hand circular polarisation, so signals with left-hand circular polarisation can be considered to be interference. Left-hand circularly polarised signals may simply be discarded, or alternatively further signal processing may be carried out such as that used in beamforming. Beamforming may be employed if there are multiple antennas and associated RF processing stages (see for example, Talwar A., ‘Interference Cancellation Improves GPS Receiver Performance’, Microwaves and RF, April 1997, pp62-68). Outputs from the antennas are combined in such a way as to counteract interference. Frequency excision filters (time-series filters or equivalents) may be used to reject interference signals in certain frequency regions (see Badke B. and Spanias AS, ‘Partial band Interference excision for GPS using frequency-domain exponents’, IEEE Proc, ICASSP'02, Vol. 4, 13-17 May 2002, pp3936-39). Data-adaptive signal processing algorithms may be used to identify which regions need to be removed and to control the filters. This approach is limited to narrow-band interference signals that only occupy a small fraction of the bandwidth of GPS signals. A further alternative uses nonlinear companding: in this approach, a GPS signal is transformed by a memoryless nonlinear function with a form determined by analysing the signal (see Przyjemski J., Balboni E. and J. Dowdle J., ‘GPS Anti-jam Enhancement Techniques’, Proc. ION, 49™ Meeting, Cambridge, June 1993, pp44-50). The objective is to reduce the power in an interference signal whilst leaving the GPS signal unaffected. This approach does not work well when the interference signal has Gaussian or nearly Gaussian statistics. In ‘Single-channel blind signal separation of filtered MPSK signals’, IEE Proc. Radar, Sonar and Navigation, Vol. 150, No. 6, Dec 2003, pp396-402, E.S. Warner and |.K. Proudler disclose a method for separating digital communications signals based on the use of time-series filters and a higher than normal sampling rate. It incorporates a generalised side lobe canceller (GSLC), which is related to a beamformer. A desired digital communications signal to be discerned has a positive signal to noise ratio (SNR), 3 i.e. the desired signal has more energy than background noise so the technique is referred to as high SNR signal separation. Unfortunately this Is not the case for GPS RF signals which have a low SNR of typically less than -20dB (GPS L1 CIA code), but also down to as low as -34 dB (typical GPS L1 Y code). These values are disclosed in NAVSTAR GPS Space Segment / Navigation User Interfaces (Public Release Version), ARINC Research Corporation, 11770 Wamer Avenue, Suite 210, Fountain Valley, CA, 92708, July 3, 1991. However, if the interference has a positive SNR one might attempt to use this approach to recover a copy of it from background noise and GPS signals. The recovered interference could then be subtracted from the received signal (interference plus noise plus GPS signal) with the intention of leaving only noise and GPS signals i.e. the situation in the interference-free case. This has been investigated, but unfortunately enough interference remained after subtraction that GPS signals could not be detected reliably using standard techniques. It is an object of the present invention to provide an alternative interference mitigation technique. The present invention provides an apparatus for mitigating interference experienced by a receiver in a global navigation satellite system (GNSS), the apparatus comprising: a) an antenna for receiving GNSS signals, b) means for oversampling signals received by the antenna to generate a series of digital signal samples, c) means for converting the series of digital signal samples into parallel data streams which are decimated versions of the series, at least some of the streams containing different sets of samples and containing between them the majority or all of the digital signal samples so that each of this majority or all appears in at least one of the parallel data streams, d) means for performing signal component analysis (SCA) upon the data streams to yield statistically distinct signals, 8) means for correlating GNSS spreading codes with the statistically distinct signals to obtain correlation scores, and f) a GNSS receiver for processing the statistically distinct signal having the highest correlation score for provision of a position fix. 4 The invention provides the advantage that it can discern and use a GNSS signal even though that signal is obscured by interference of much higher power. In a computer- greater than that of the GPS signals was successfully removed. Further advantages are that it can mitigate a wide range of interference types and can use but does not require multiple antennas. The means for performing SCA may be a means for performing independent component analysis (ICA); which may in turn incorporate a principal component analysis (PCA) stage which produces uncorrelated signals followed by a discard stage for discarding some uncorrelated signals and a Higher Order Statistics (HOS) stage for processing undiscarded uncorrelated signals. The discard stage may be arranged to discard uncorrelated signals that are intermediate to strong and weak. and do and do not contribute significantly to PCA stage input signals respectively, and to relay to the HOS stage uncorrelated signals contributing marginally to PCA stage input signals, contributions being assessed as intermediate to strong, marginal or weak in accordance with whether they are associated respectively with normalised eigenvalues or normalised singular values that are above, between or below two thresholds in a normalised distribution of eigenvalues or singular values produced by the PCA stage. The PCA stage may be arranged to associate each of the uncorrelated signals with a respective energy level representing how much of a contribution the uncorrelated signal makes to PCA stage input signals, and to derive the thresholds by differencing the energy levels. The energy levels may be obtained from normalised eigenvalues or singular values either directly or after information theoretic processing. The PCA stage may be arranged to derive the thresholds by ordering the energy levels so that their values change monotonically, to calculate first differences between adjacent pairs of ordered energy levels, to calculate second differences between adjacent pairs of first differences, and to obtain an intermediate to strong-marginal boundary from a first maximum in the second differences progressing from higher to lower energy levels, and to obtain a marginal-weak boundary from a first maximum to occur in the first differences after a first minimum in the first differences. The apparatus of the invention may be for use with a GNSS signal having a spreading code with a chip rate, each parallel data stream consisting of samples at a rate that is at least twice the chip rate. The means for converting the series of digital signal samples into parallel data streams may be arranged to implement a decimation rate of at least 3, and to produce at least 3 parallel data streams. The means for oversampling signals may be arranged to oversample at a rate in the range 4Rc to 20R¢, preferably 12R¢, where Re is a GNSS chip code rate. In another aspect, the present invention provides a method for mitigating interference experienced by a GNSS receiver, the method having the steps of: a) using an antenna to receive GNSS signals, b) oversampling signals received by the antenna to generate a series of digital signal samples, c) converting the series of digital signal samples into parallel data streams which are decimated versions of the series, at least some of the streams containing different sets of samples and containing between them the majority or all of the digital signal samples so that each of this majority or all appears in at least one of the parallel data streams, d) performing signal component analysis upon the data streams to yield statistically distinct signals, e) correlating GNSS spreading codes with the statistically distinct signals to obtain correlation scores, and f) using a GNSS receiver to process the statistically distinct signal having the highest correlation score for provision of a position fix. The method aspect of the invention may have preferred features equivalent mutatis mutandis to those of the apparatus aspect. In order that the invention might be more fully understood, embodiments thereof will now be described by way of example only, with reference to the accompanying drawings, in which: Figure 1 is a schematic block diagram of a device of the invention for single channel © interference mitigation for GPS signals; and Figure 2 is a graph of normalised singular values against singular value index which is used to set boundaries to separate GPS signals from interference and noise. Before describing the drawings, the term ‘oversampling’ will be defined. It is well known that a continuous time, band-limited signal must be sampled at a rate equal to at least twice the highest frequency present in the signal (the Nyquist rate) to avoid information being lost in a sampling process. If the signal is strictly band-limited, then sampling at a rate commensurate with the passband edges and the signal bandwidth is known to be sufficient and is implicitly taken as the Nyquist rate (see Gibson J.D., editor-in-chief, ‘The Mobile Communications Handbook’, pub. CRC Press, 1996). Sampling faster than the Nyquist rate is referred to as ‘oversampling: it is usually considered to be unnecessary to sample significantly faster than Nyquist as it can lead to devices that are more power hungry and costly than is conventionally considered necessary and few benefits if any are known. In specific applications, e.g. analogue-to-digital conversion, oversampling has been shown to have benefits but in connection with other non-standard methodology such as one bit sampling. The embodiment of the invention to be described relates to mitigation of interference in a GPS receiver. However, the invention can be used with receivers of other global navigation satellite systems. Referring to Figure 1, an interference mitigation device indicated generally by 10 incorporates an antenna 12 which picks up GPS RF signals and interference. The antenna 12 is a single antenna in this example, although multiple antennas may be used. The GPS RF signals and interference are processed by an appropriate RF / IF front end 14, IF being an intermediate frequency to which the RF signals are down converted by mixing with a local oscillator signal in a conventional way. A signal output from the front end 14 is digitised by an analogue to digital converter (ADC) 16, which oversamples it: i.e. the ADC 16 samples the signal faster than the GPS C / A chip code rate. This oversampling is at a rate which is determined by the number of interference sources present. For one interference source an oversampling rate of 12R. has been found to ‘work well, where R; is the GPS C / A chip code rate of 1.023Mcps, and the oversampling rate may be increased above or reduced below this. This example of the invention should also work acceptably with L1 and L2 Y-code signals with a chip code rate of 10.23Mcps. The ADC 16 digitises into at least 8 or 9 bits, as opposed to 1 or 2 bits normally used in civil GPS C / A code receivers. The oversampled signal output from the ADC 16 is passed to an upper end of a chain of (M-1) time delay elements 18 imposing equal time delays: of the elements 18, pr uppermost and lowermost elements 18, and 18y. are illustrated and intervening elements (if any) are indicated by a chain line 20. The ADC output is undelayed upon an uppermost line 22;, but successive lines 22; to 22y receive the ADC output with progressively increasing delay: i.e. line 22, receives the ADC output with delay (p-1)5, where p=1to M, § is the delay imposed by each of the time délay elements 18 and is equal to the time delay between sampling instants of the ADC 16. Each of the lines 224 to 22y is connected to a respective decimation operator 24; to 24y; each decimation operator 24 has the property that it outputs only one in every N of the signal samples it inputs, N being referred to as the decimation rate. Thus, at the outputs of the decimation operators 24, the output of the ADC 16 has become converted into a plurality of streams of samples in parallel with relative delay and with a sample rate which is 1 / N of that of the ADC. A different set of samples is extracted for some or all of the decimation operator output signals so that each of the samples from the ADC output signal appears in at least one of the decimation operator output signals. It is preferable for all of the ADC output samples to appear in one of more of the decimation operator output signals, but loss of a small proportion of the samples (e.g. 5% of the total number of samples) is not too serious. The decimation rate N and the number of parallel signals M may be independent of each other, but it is preferred that they are equal to one another. In the case of one interfering signal, the use of M = 3 parallel signals and a decimation rate of N=3 has been found to work well, but they may be greater than 3. If however M and N have values which are different to one another, M<N implies that not all the ADC output samples are present in the M parallel signals, and MN implies that some ADC output samples are present more than once in the M parallel signals. Output signals from each decimation operator 24 are sampled at a rate which is at least twice a GPS signal's spreading code chip rate. The M decimation operator output signals are each passed as an input signal to a respective tapped delay line 28, to 28y. The output from a tapped delay line (TDL) is a set of P signals that consist of the input signal and (P-1) delayed versions of the input signal. The set of (P-1) delayed signals is such that for each integer between 1 and (P-1) inclusive there is a signal in the set that is delayed by a number of samples equal to that integer. In this embodiment P = 12, but P can be greater or smaller than this. The P output signals from each of the M TDLs 28 are fed into a blind signal (or source) separation stage 30, which implements Independent Component Analysis (ICA). There are several blind signal separation algorithms in the open literature, which Is reviewed in R WO 03 / 073812. This embodiment uses a type of blind signal separation that separates the signals in three steps: a Principal Component Analysis (PCA) stage 32 which generates signals that are uncorrelated: this is followed by ‘a discard stage 34 that discards some PCA stage output signals and a Higher Order Statistics (HOS) stage 36. This type of blind signal separation using PCA followed by HOS is known — see e.g. WO 2005 / 052848. ICA and PCA are special cases of Signal Component Analysis (SCA). PCA can be used on its own without HOS stage 36, but the addition of the HOS stage improves results significantly. The PCA stage 32, which uses second order statistics, decorrelates and normalises its input signals. The HOS stage 36, which normally relies on fourth order statistics (kurtosis), makes the independent normalised signals (obtained from the PCA stage 32) independent and thus statistically distinct from one another: here “statistically distinct” means at least one of uncorrelated and independent. The discard stage 34 is used in an unconventional manner in this embodiment. The PCA stage 32 carries out a singular value decomposition or SVD of the input signals (see Haykin S., ‘Adaptive Filter Theory’, 2", Ed., pub. Prentice-Hall, Englewood Cliffs, New Jersey, 1991): this transforms the input signals into a set of signals (plus noise) that are mutually uncorrelated and normalised to unit signal energy. In addition, the PCA stage 32 associates each of the uncorrelated signals with a quantity (referred to as the ‘energy level’) which is a respective singular value representing how much of a contribution the uncorrelated signal makes to PCA stage input signals. Usually in blind signal separation, any uncorrelated signal that is weak and does not contribute significantly to the PCA stage input signals is discarded before the HOS stage. In this example, uncorrelated signals that are intermediate to strong and contribute significantly to the PCA stage input signals are also discarded. Consequently, only uncorrelated signals that make a small but above noise contribution to the PCA stage input signals are retained (hereinafter referred to as a marginal contribution or marginal signals). The reason for this is that uncorrelated signals that make a marginal contribution to the PCA stage input signals contain GPS signals and noise, whereas uncorrelated signals that are intermediate to strong and contribute significantly in this regard correspond to interference plus noise, and those which do not so contribute correspond to noise alone. To select the marginal signals for retention, the energy levels are processed: they may be processed directly or after applying information-theoretic processing to them, e.g. the Akaike information criterion, or AIC: see Akaike H., ‘A new look at the statistical model identification’, IEEE Trans. Autom. Control, Vol. AC-19, No. 6, Dec 1974, pp716-23; an 9 alternative information-theoretic processing technique is referred to as the minimum description length (MDL) technique: see Wax M. and Kailath T., ‘Detection of Signals by Information Theoretic Criteria’, IEEE Trans. Acoustics, Speech and Signal Proc., Vol. ASSP-33, No.2, Apr 1985, pp387-92). Both AIC and MDL are also given and referenced in the Haykin reference mentioned above: in either case, information-theoretic processing converts the energy levels from the PCA stage into a set of numbers that can also be thought of as energy levels. The signals for retention are selected by determining two boundaries. In this embodiment a preferred method of obtaining the boundaries is based on differencing the energy levels calculated by the PCA stage 32 after applying the MDL method mentioned above. This method is implemented by an extension to the PCA stage 32. The PCA energy levels are ordered so that their values are in a monotonically decreasing sequence: consequently, in the following description, in the sequence “before” and "after’ mean at higher and lower energy levels respectively. First differences between adjacent pairs of ordered energy levels are calculated. Then second differences between adjacent pairs of first differences are also calculated. The boundaries or thresholds between the intermediate to strong, marginal and weak uncorrelated signals described above are calculated as follows. Maxima are observed in the first and second differences: the index of a first maximum in the second differences sequence is used to indicate the energy level lower boundary between the intermediate to strong and marginal uncorrelated signals (here “lower” indicates that this maximum occurs at a lower singular value index, not that the singular value (or energy level) is lower, and “first’ is in the context of moving from higher to lower energies). A first minimum in the first differences sequence serves to locate a maximum that indicates the singular value index boundary between the marginal and weak uncorrelated signals: this maximum is the first to occur after the minimum. Energy levels could instead be ordered so as to increase monotonically, in which case the above maxima and minima would be determined from energy levels in reverse order. Although in this embodiment, boundaries or thresholds are derived from singular values, eigenvalues may be used in a similar manner, or approximations to either of these. Referring now also to Figure 2, PCA stage results are shown as normalised singular values plotted against singular value index. These results are used to set boundaries (or thresholds) to separate GPS signals from interference and noise: here normalised singular values above an upper threshold 50 of about 1.4 correspond to intermediate to strong uncorrelated signals. Marginal uncorrelated signals correspond to normalised singular values below the upper threshold 50 but above a lower threshold 52 of about 0.01. Weak uncorrelated signals correspond to normalised singular values below the lower threshold 52. The HOS stage 36 transforms the marginal uncorrelated signals into ICA signals for output to a detector stage 38, which correlates each ICA signal with known GPS spreading codes: here a semi-coherent correlator is used rather that a standard non- coherent correlator. Each correlation process produces a positive correlation number indicating the likelihood that the ICA signal contains the relevant GPS signal. For each ICA signal, the correlation numbers for all the GPS signals are added to produce an overall score. It is also possible to form an overall score in other ways, e.g. by summing only those correlation numbers above a predetermined threshold. The ICA signal with the largest overall score is output from the device 10 at 40 and then processed by a standard GPS receiver (not shown) in order to obtain a position fix. It is a feature of this embodiment that the ICA signal with the largest score is processed directly by a standard GPS receiver: i.e. unlike high SNR signal separation a generalised side lobe canceller (GSLC) is not required. The processing described above may be extended by using multiple antennas 12. To test the invention, a simulation was carried out which used real GPS data (which ‘included receiver noise) and an artificial interference source. The GPS data was real bandpass data sampled at 16.3676 MHz and centred at an intermediate frequency (IF) of 4.1304 MHz. The signal was down-converted to zero IF to provide complex baseband 1% signals and resampled to adjust the sample rate to 12 times the GPS C / A chip code rate. A filter was used to band-limit the signal (and receiver noise). Similarly, a band-limited BPSK interference source was generated with a commensurate sample rate and with a bandwidth which was an adjustable fraction of the GPS chip rate. The BPSK interference signal was added to the resampled, band-limited GPS data and noise to represent the received signal. Sufficient data was produced to generate a received data signal of length corresponding to 500 chip periods. For the simulation the artificial interference source had an interference-to-noise ratio (INR) of 30 dB, a bit rate of 746 kHz (i.e. fractional bandwidth (BW) of 0.73 times the chip rate of 1.023 MHz), no frequency offset and a random phase offset. 11 To mitigate the interference, the composite, oversampled, received signal was passed through a chain of three delay elements, times three down-converters, and tapped delay lines (12 taps) and delay line outputs were processed using a blind signal separation algorithm referred to as the BLISS algorithm (see Clarke 1.J., ‘Direct Exploitation of Non- Gaussianity as a Discriminant, EUSIPCO 1998, September 1998). Outputs obtained from BLISS processing were then passed through a GPS-like correlator, i.e. a semi- coherent detector (see ‘Lin D.M. and Tsui J.B.Y., ‘A Software GPS Receiver for Weak Signals’, IEEE MTT-S Intl. Microwave Symp. Digest, Vol.3, 20-25 May 2001, pp2139-42). For this detector, each 20 ms period was divided into two 10 ms periods such that within one of them there would be no data bit transition. Within each 10 ms period, conventional correlation with the spreading code was performed. The two 10 ms correlations were performed over a number of different frequencies commensurate with the expected frequency uncertainty and Doppler frequency offset. The resolution of the frequency search was needed to maintain coherency across the 10 ms period (frequency spacing <50 Hz). Table 1 below summarises preliminary simulation results obtained using real GPS data and artificial interference. Detection values are all normalised in the conventional manner for GPS detection, i.e. divided by a mean value derived from the correlation process. After interference suppression processing, GPS satellite vehicles (SVs) 3 and 22 were detected (and possibly 15 and 19), although about 10 dB down on that in the absence of interference. Due to there being different distances, or ranges, from the receiver to different satellites, there are different recorded time instants for detection occurring in different satellites. These time delays are central to the ability of a GPS receiver to obtain a position fix. It is thus important to determine whether or not the interference mitigation processing adversely affects these time delays. Absolute time delays can only be determined by processing as employed by a GPS receiver. In the absence of such processing, a simple alternative is to choose a one satellite as a reference and look at the difference in time delay between each of the other satellites and the reference satellite. Taking GPS SV 3 as a reference and subtracting this delay from those for GPS SVs 15, 19 and 22, Table 2 shows expected relative time delays (in units of one-quarter of a C / A code chip) averaged over fifty 20 ms periods. Differences between means for no interference and with interference and processing were calculated: these are shown in Table 2, and corresponded to range differences of 39.6, 5.6 and 0.6m. In this example, interference suppression processing was successful and did not seem to perturb the relative time differences unduly. 12 GPS satellite vehicle Detection value (no Detection value (with Ee mn El] Yi ta eX a number interference) in dB interference, after pa processing) | 3 26.95 16.55 LIL Vibe fel SWRI wR aes arference) in dB interference, after J processing) in dB 26.95 16.56 1014 v 14 B89 11 rT 18.14 1 11.52 14 | 16.55 | 11.29 — 15 | 24.83 | 12.58 16 | 22 57 | 11.12 18 1 23.19 7 10.85 19 25.38 | 12.28 11 18.14 11.52 3 14 16.55 11.29 — 15 24.83 12.59 16 22.57 11.12 18 23.19 10.85 19 25.38 12.28 21 16.83 10.41 22 26.92 16.57 Table 1: Simulation Results for a real GPS data and artificial jamming. 398 5.6 GPS SVs | Relative time-delay, | Relative time delay, | Difference | Equivalent no interference with interference, in means range after processing error (m 39.6 5.6 0.6 V22-8V3 | 1676.8942+ 0.7268 | 1676.9017+1.2414 | -0.0075 0.6 Table 2: Example of relative time-delay estimates (In units of a quarter of a C / A code chip) with real GPS data.

Claims

Apparatus for mitigating interference experienced by a global navigation satellite system (GNSS) receiver, the apparatus comprising:a) an antenna for receiving GNSS signals,b) means for oversampling signals received by the antenna to generate a series of digital signal samples,c) means for converting the series of digital signal samples into parallel data streams which are decimated versions of the series, at least some of the streams containing different sets of samples and containing between them the majority or all of the digital signal samples so that each of this majority or all appears in at least one of the parallel data streams,d) means for performing signal component analysis (SCA) upon the data streams to yield statistically distinct signals,e) means for correlating GNSS spreading codes with the statistically distinct signals to obtain correlation scores, andf) a GNSS receiver for processing the statistically distinct signal having the highest correlation score for provision of a position fix.

2. Apparatus according to claim 1 wherein the means for performing SCA is a means for performing independent component analysis (ICA).

3. Apparatus according to claim 2 wherein the means for performing ICA incorporates a principal component analysis (PCA) stage arranged to produce uncorrelated signals followed a discard stage for discarding some uncorrelated signals and a Higher Order Statistics (HOS) stage for processing undiscarded uncorrelated signals.

4. Apparatus according to claim 3 wherein the PCA stage is arranged to generate singular values or eigenvalues and to modify them by information theoretic processing.

5. Apparatus according to claim 4 where the information theoretic processing is that associated with a technique referred to as the minimum description length method.

6. Apparatus according to claim 5 wherein the PC A stage is arranged to produce a . normalised distribution of singular values or eigenvalues, and the discard stage is arranged to discard uncorrelated signals that are intermediate to strong and weak and do and do not contribute significantly to PCA stage input signals respectively, and to relay to the HOS stage uncorrelated signals contributing marginally to PCA stage input signals, contributions being assessed as intermediate to strong, marginal or weak in accordance with whether they are associated respectively with singular values or eigenvalues that are above, between or below two thresholds in the normalised distribution.

7. Apparatus according to claim 6 wherein the PCA stage is arranged to associate each of the uncorrelated signals with a respective energy level representing how much of a contribution the uncorrelated signal makes to PCA stage input signals, and to derive the thresholds by differencing the energy levels.

8. Apparatus according to claim 7 wherein the PCA stage is arranged to derive the thresholds by ordering the energy levels so that their values change monotonically, to calculate first differences between adjacent pairs of ordered energy levels, to calculate second differences between adjacent pairs of first differences, and to obtain an intermediate to strong-marginal boundary from the first maximum in the second differences , and to obtain a marginal-weak boundary from the first maximum in the first differences occurring after the first minimum in the first differences.

9. Apparatus according to claim 1 wherein the means for correlating GNSS spreading codes is arranged to obtain a value for the correlation of each statistically distinct signal with each spreading code and to sum the correlation values for each statistically distinct signal to obtain a correlation score for that signal.

10. Apparatus according to claim 1 for use with a GNSS signal having a spreading code with a chip rate, and each parallel data stream consists of samples at a rate that is at least twice the chip rate.

11. Apparatus according to claim 1 wherein the means for converting the series of digital signal samples into parallel data streams is arranged to implement a decimation rate of at least 3, and to produce at least 3 parallel data streams.

12. Apparatus according to claim 1 wherein the means for oversampling signals is arranged to oversample at a rate in the range 4RC to 20Rc, where Rc is a GPS chip code rate.

13. A method for mitigating interference experienced by a GNSS receiver, the method having the steps of:a) using an antenna to receive GNSS signals,b) oversampling signals received by the antenna to generate a series of digital signal samples,c) converting the series of digital signal samples into parallel data streams which are decimated versions of the series, at least some of the streams containing different sets of samples and containing between them the majority or all of the digital signal samples so that each of this majority or all appears in at least one of the parallel data streams,d) performing signal component analysis (SCA) upon the data streams to yield statistically distinct signals,e) correlating GNSS spreading codes with the statistically distinct signals to obtain correlation scores, andf) using a GNSS receiver to process the statistically distinct signal having the highest correlation score for provision of a position fix.

14. A method according to claim 13 wherein the step of performing SCA is implemented by performing independent component analysis (ICA) and wherein the GNSS signals have a signal to noise ratio of less than OdB.

15. A method according to claim 14 wherein the step of performing ICA incorporates principal component analysis (PCA) producing uncorrelated signals followed by a step of discarding some uncorrelated signals and a step of performing Higher Order Statistics (HOS) upon undiscarded uncorrelated signals.

16. A method according to claim 15 wherein the step of performing ICA generate singular values or eigenvalues modified by information theoretic processing .

17. A method according to claim 16 wherein the information theoretic processing is that associated with a technique referred to as the minimum description length method.

18. A method according to claim 15 wherein the discarding step discards uncorrelated signals that are intermediate to strong and weak and do and do not contribute significantly to PCA input signals respectively, and the HOS step processes uncorrelated signals contributing marginally to PCA input signals, contributions being assessed as intermediate to strong, marginal or weak in accordance with whether they are associated respectively with normalised singular values that are above, between or below two thresholds in a normalised singular value distribution produced by the PCA step.

19. A method according to claim 14 wherein the PCA step associates each of the uncorrelated signals with a respective normalised singular value representing how much of a contribution an uncorrelated signal makes to PCA input signals, and derives the thresholds by differencing the energy levels.

20. A method according to claim 15 wherein the PCA step derives the thresholds by ordering the energy levels so that their values change monotonically, calculating first differences between adjacent pairs of ordered energy levels, calculating second differences between adjacent pairs of first differences, obtains an intermediate to strong-marginal boundary from a first maximum in the second differences , and obtains a marginal-weak boundary from a first maximum in the first differences occurring after a first minimum in the first differences.

21. A method according to claim 13 wherein each GNSS signal has a spreading code with a chip rate, and each parallel data stream consisting of samples at a rate that is at least twice the chip rate.

22. A method according to claim 13 wherein the step of converting the series of digital signal samples into parallel data streams implements a decimation rate of at least 3, and produces at least 3 parallel data streams.

23. A method according to claim 13 wherein the step of oversampling signals oversamples at a rate in the range 4RC to 20Rc where Rc is a GNSS chip code rate.

Images