WO2007031088A1

WO2007031088A1 - A method for sending secure information or increasing communication capacity via coding of wavefronts and a system using said method

Info

Publication number: WO2007031088A1
Application number: PCT/DK2006/000508
Authority: WO
Inventors: Patrick Eggers; Petar Popovski; Xin Zhou; Frank Fitzek; Persefoni Kyritsi
Original assignee: Aalborg Universited
Priority date: 2005-09-15
Filing date: 2006-09-15
Publication date: 2007-03-22

Abstract

The invention relates to a method for sending secure information or increasing communication capacity in a communication system comprising a transmitter side and a receiver side, said receiver side having a number of sensors and said transmitter having a number of emitters. The transmitter side first transmits a data stream containing information in such a way that a certain spatial sensor response is produced at the sensors on the receiver side, the sensor response carrying at least part of the information. Subsequently, the receiver side detects and interprets the spatial sensor response and thereby extracts at least the previously mentioned part of the information.

Description

Title: A method for sending secure information or increasing communication capacity via coding of wavefronts and a system using said method

Technical Field

The present invention relates to a method for sending information in a communication system comprising a transmitter side and a receiver side, said receiver side having a number of sensors and said transmitter side having a number of emitters. The invention further relates to a communication system adapted to utilize the method for send- ing information.

Background Art

In wireless communication system design, the main objectives are system efficiency improvement, link reliability enhancement, and security of transmission. The system efficiency is reflected in the channel capacity, which refers to the maximum possible information transfer rate of the channel [1], while the link reliability is indicated by the signal quality [2].

Wireless data communication is commonly coded in the time or frequency domain. The interaction of the sensor array and the transmitted signal is manipulated solely for the purpose of maximizing throughput/quality/capacity, regardless of the resulting particular signal state at each individual sensor, i.e. it is the combination that is optimized. The data should be decoded by the intended user but can possibly be decoded by other users at neighboring locations or with neighboring sensor arrays. Therefore, most security measures in wireless communications are performed via secret keys or coding at higher layers. Thereby, eavesdroppers can listen in on the uncoded information but cannot successfully decode it without knowledge of the appropriate key.

Pioneering studies have shown that the utilization of multiple antennas at both the transmitter (Tx) and the receiver (Rx) ends can achieve high capacity [7] [8]. A way to realize this is by using spatial multiplexing along orthogonal parallel channels given by the Singular Value Decomposition (SVD) processing technique [9]. Each orthogonal channel is defined as a transmitter and receiver array pattern pair. Using this transmis- sion technique, each data channel is interference-free (see Figure 1). However, this method requires high computational complexity at both the Tx and Rx sides. In addi- tion, the number of maximum orthogonal channels is in theory limited by the minimum number of Tx antennas M and Rx antennas N, namely min(M,N) with min(-) denoting the minimum value. However, the effective number of orthogonal channels is smaller than min(M,N), since the subchannels do not all have equal gains.

According to the SVD technique, min(M,N) orthogonal pairs of Tx-Rx patterns can be created. If the orthogonality constraint is relaxed, then a much higher number of patterns can be created. Non-orthogonal channels have the following drawback: although the main beam of the antenna pattern focuses on the intended Rx antennas, some of the side lobe beams may point to the other Rx antennas and cause a leakage signal. Such a leakage signal acts as interference, which results in reduction of the signal quality and capacity. Clearly, there is a trade-off between the number of channels and the capacity performance. Overlaying can be used on both orthogonal communication systems and non-orthogonal communications.

The object of the invention is to: a) provide a method for secure overlay communication via spatially referenced waveform coding, and b) improve the capacity, signal quality and spectrum efficiency for a communication system. Utilizing overlay communication can add to the effective modulation order of the complete communication system.

If the waveform features are changed across the sensor array (phase, amplitude etc), additional overlay data can be transmitted. Overlay data should be decoded by the intended user and must not be decodable by any other user at a different location and/or sensors arrays. By manipulating the waveform at a particular receiver array and loca- tion, the overlay data is inherently secure, since its decoding is based on spatially referenced changes in the signal properties, such as phase, amplitude, delay, frequency, and polarization state (tilt/ellipticity etc.) in the case of radio communication. Thereby, physical layer security is achieved, regardless of whether any higher layer coding is present or not. Here the overlay information bearing part is the particular signal state at one sensor as compared to the signal state at other sensor(s).

Disclosure of Invention

This is according to the invention achieved by a method for sending information in a communication system comprising a transmitter side and a receiver side, said receiver side having a number of sensors and said transmitter having a number of emitters, wherein a) the transmitter side transmits a data stream containing information, which is divided into a first part and a remaining part, in such a way that a certain spatial sensor response is produced at said number of sensors on the receiver side, said sensor response carrying at least the first part of said information contained in the data stream, b) the receiver side detects and interprets said spatial sensor response and extracts at least the first part of said information, and c) the receiver side detects and interprets the remaining part of the information.

Thereby, part of the information can be conveyed as a spatial sensor response by coding the propagated waveform. In other words: 'the medium is the message'. The medium waveform response is particular to the array and locations of the acting transmitter and the acting receiver. Therefore, physically (or spatially) secure communication is inherently achieved. Furthermore, it is possible to increase the total throughput, as the overlay data increases the effective total modulation order of the complete communication system (i.e. total bits/symbol time). Thus overlay data also provides better flexibility to approach theoretical capacity limits.

The communication can be divided into a secure overlay data rate in full parallel with base or bulk information. The overlay information is useful for simultaneous conveying of decryption keys (to decode the base information), account information or similar sensible data, which do not require a high throughput. The overlay information is coded directly on the received waveform and is thus particular for only the intended transmit- ter/receiver pair, and it is not possible for an eavesdropper to intervene due to unavoidable differences in the physical location of the spatial sensor responses of the receiver and eavesdropper, respectively.

According to a particular embodiment of the invention, the sensor response comprises physical waveform samples across a sensor array, retrieved at specific sensor locations at the receiver side. This ensures overlay capability on physically coded Multiple Input Multiple Output (MIMO) schemes.

According to another particular embodiment, the sensor response comprises trans- formed waveform samples in a transformed eigen domain, whereby each created channel or eigen mode is acting as a virtual sensor. This ensures overlay capability in orthogonal eigen mode MIMO schemes.

According to a preferred embodiment of the invention, the data stream at the transmit- ter side is divided into a base part and an overlaid part, and the receiver side decodes the base part using standard decoding methods and decodes the overlaid part using the sensor response detected. The base part corresponds to communication, which per se is known from conventional modem communication, while the overlaid part corresponds to the waveform coded part of the information or the aforementioned first part of the data stream.

In a particularly preferred embodiment according to the invention the transmitter side deliberately produces said spatial sensor response at the receiver side to reflect at least the first part of said information. In other words the transmitter side produces the desired coded waveform based on the first part of said information, which is transmitted to the receiver side and subsequently decoded. The sensor response is interpreted with respect to signal characteristics, such as amplitude, phase, polarization, delay or frequency, of the signals detected at the receiver sensors. Thereby, a simple method for interpreting the coded waveform at the receiver side is provided.

According to another preferred embodiment at least the first part of said information is coded using differential sensor responses between two or more individual sensors of said number of receiver sensors.

According to yet another preferred embodiment, the receiver side uses interpretation information, such as an overlay coding table or a modulation transfer function, so as to extract at least the first part or the overlaid part of the information based on the sensor response received at the receiver side.

In a particular embodiment according to the invention, the interpretation information at the receiver side is dynamically updated. Thereby, the system can automatically correct for changes in the transmission conditions.

According to a preferred embodiment of the invention, the sensor response at the re- ceiver side is digitized with respect to signal characteristics, such as any combination of amplitude, phase, polarization, delay or frequency. According to another preferred embodiment of the invention, the sensor response at the receiver side is digitized using a number of decision regions. For 2-dimensional modulations like Quadrature Amplitude Modulation (QAM) the decision regions are in the complex plane. For one dimensional modulations, the decision regions are determined by threshold levels. For example, one threshold level is used for producing a 1- bit symbol, while three amplitude threshold levels are needed for producing a 2-bit amplitude coded symbol, etc. In general the number of amplitude threshold levels should be 2^symbols-1 , but if phase/complex thresholds are used (like for QPSK-like constella- tions) fewer thresholds are needed (only two for a 2-bit QPSK constellation).

In a preferred embodiment according to the invention, each of the received signals at said number of sensors at the receiver side is compared to said number of threshold levels. Thus, each sensor produces a symbol containing a number of bits. The total bits from all the sensors constitute the overlaid part of the data stream.

In another preferred embodiment according to the invention, the digitisation conditions are dynamically exchanged between the transmitter side and the receiver side. Thereby, the thresholds are updated in needed intervals with respect to changes in transmission conditions, such as the channel state information, interference and/or noise.

In another preferred embodiment according to the invention, a time reversal transmission scheme uses non-orthogonal parallel channels for sending the data stream from the transmitter side to the receiver side.

According to a preferred embodiment, the transmitter side prior to step a) estimates an impulse response h(t) for a channel between the transmitter side and a given receiver sensor and subsequently generates a filter weight for the channel, said filter weight be- ing the time reversed conjugate h*(-t) of the estimated impulse response, and wherein the transmitter side subsequently convolves the data stream with the filter weight and in step a) transmits the resulting convolved signal to the receiver. Thus the received signal is the convolution of the input data stream convolved with the autocorrelation function of the channel impulse response. This function has the property that the maximum value is found at t = 0 and corresponds to the power of the input channel. This technique is used for the base communication. Due to the spatial focusing ability on a single receiver sensor (i.e. the possibility to achieve a high/max amplitude focus on one receiver sensor compared to the other receiver sensors) it is possible to interpret the outcome as a data word coding along the sensor array.

According to a preferred embodiment, at each sensor at the receiver side in step b) the received signal is given as the sum of the signals transmitted from the transmitter elements. Each of the signals appears as an autocorrelation function. Thereby, the side lobes of the autocorrelation functions are inherently suppressed, due to coherent addition of the main peaks and non-coherent summation of the side lobes.

According to a preferred embodiment, the method is applied in a multiple input multiple output (MIMO) system, in which the transmitter side has a number of transmitter elements, a number of channels being generated between said transmitter elements and said receiver sensors, and wherein the transmitter side prior to step a) for each trans- mitter element estimates an impulse response for each channel h_m(t) between said transmitter elements and the individual receiver sensors and subsequently generates a filter weight, wherein the transmitter side subsequently convolves the data stream with each of the filter weights for the individual transmitter elements, and in step a) transmits the resulting convolved signal to the receiver.

According to a preferred embodiment, a time reversal scheme is used for weighting of an emitter array on the transmitter side, so that a higher weighting is applied to those emitters, which have a highest instantaneous differential link gain to the intended receiver sensor compared to the remaining emitters.

According to another preferred embodiment, a time reversal transmission scheme is used for coding the transmitted data stream to produce a pulse/time modulated waveform, such as a pulse-position modulated (PPM) waveform, overlaying the sensor array. This embodiment is particularly applicable due to the temporal focusing ability on single sensors in ultra wideband (UWB) applications, i.e. pulse shifting on one sensor compared to the other sensors, and which is interpreted as an outcome of a data word coding along the sensor array.

According to an alternative embodiment, a polarisation state transmission scheme is used for coding the transmitted data stream to produce polarisation modulated waveform overlaying the sensor array. For instance binary left versus right hand polarisation can be used, or for finer resolution using polarisation tilt and/or elliptically coded polarization. The degree of resolution to use on polarisation coding depends on the degree of polarisation discrimination possible in the radio environment and/or potential for creating perceived polarisation purity at the sensor array to be used for conveying informa- tion.

According to a preferred embodiment, the overlay information is coded in a higher order version of the base communication modulation scheme, such as digital phase, amplitude or frequency modulation, such as ASK, QAM, PSK, or FSK. For example, if the base communication is performed with binary phase shift keying (BPSK), higher order phase shift keying states can be used for overlay. This is possible in a low rate overlay block coding, where the higher (via averaging gain) block SNR allows for finer granularity of detection/states than the one possible for the base communication. If base communication for example is binary frequency shift keying (BFSK) modulated, then higher order FSK overlay states can be used with fractional frequency shifts (of the frequency shift representing the difference of base communication '0' and '1' states). This overlay frequency shift (at block level) can for example be extracted from the frequency synchronisation loops of the receiver.

The object according to the invention is also achieved by a communication system for sending information, said system comprising a transmitter side and a receiver side, wherein said transmitter side comprises transmitter means for transmitting a data stream containing information and being adapted to transmit the data stream with a spatial pattern, which is at least dependent on a first part of said information, and said receiver side comprises a number of sensors, which are adapted to detect a spatial sensor response, said receiver side being adapted to interpret the spatial sensor response as information, so as to extract at least said first part of the information from said data stream. Thereby a communication system, which is able to convey information as a spatial response by coding the propagated waveform, is provided.

Preferably, said transmitter side comprises one or more transmitter elements and said receiver side comprises one or more receiver sensors. According to the invention, the transmitter side splits the data stream up into a base part and an overlaid part. The transmitter can transmit a data stream containing information. The receiver side com- prises decoding means known per se for extracting the base part of the data stream. Moreover, the transmitter can be adapted to transmit the data stream with a spatial pat- tern, the choice of which depends on said overlay information. The receiver side comprises a number of sensors, which detect the spatial sensor response, said receiver side being able to interpret the spatial sensor response as information by means of for example an overlaying table, so as to extract said first part of the information from said data stream. The sensor response can be digitised using a number of decision regions in order to produce symbols having a given number of bits for each receiver sensor. Thereby, the total information from the data stream can be extracted. Hence, a communication system, which is able to convey information as a spatial response by coding the propagated waveform on top of conventional data transmission, is provided.

The properties of the received wavefront that can be sensed are amplitude, phase, polarization, delay, and/ or frequency. In the case of radio communications, the properties of the wavefront are sensed by radio transceivers with antennas. The same techniques can be applied to other types of waves, by sensing a subset of the above properties. Alternatively, the transmitter comprises acoustic transducers (like loudspeakers, vibrators etc) and the receiver sensors are also acoustic transducers (like microphones, hydrophones etc).

According to another preferred embodiment, the transmitter side is adapted to dynami- cally update the transmission conditions in order to produce the desired spatial sensor response on the receiver side based on said first part of the information. Thereby, the communication system can automatically correct for changes in the transmission conditions.

According to a preferred embodiment, the receiver side additionally comprises decoding means known per se for extracting the data part of the data stream. Thereby, the total information from the data stream can be extracted.

According to another preferred embodiment, the system is a multiple input multiple output (MIMO) system, said transmitter side having a plurality of transmitter elements, and said receiver side having a plurality of receiver sensors, a number of orthogonal or non-orthogonal channels being generated between said transmitter elements and receiver sensors.

Preferably, the transmitter side additionally comprises filter weighting means for weighting the data streams transmitted from the transmitter element. According to a particular embodiment, the weighting means are adapted to weight the data streams according to a time reversal scheme. Alternatively, the time reversal scheme can be used for selective weighting of an emitter array on the transmitter side, so that a higher weighting is applied to those emitters, which have a highest instantaneous differential link gain to the intended receiver sensor compared to the remaining emitters. However, this scheme is not very efficient as it only controls one sensor response at the time (the waveform has just one anchor/focus point and the rest of the sensors provide uncontrolled signal (interpreted as waveform or spatial noise)).

All the aforementioned methods and embodiments can be expanded for use in a multi datastream MIMO system. For non-orthogonal space multiplexing, as for instance explained in US provisional application No. 60/791,984 by the same applicant, the transmitter weights seek to maximize receiver sensor isolations (as each data stream is associated to one receiver sensor). Consequently, this scheme is also an implementation of simultaneous and individual receiver sensor state (waveform) control, allowing best possible overlay coding.

Brief Description of the Drawings

The invention is explained in detail below with reference to the drawings, in which

Fig. 1 shows an orthogonal channel diagram of a prior art MIMO system,

Fig. 2 shows a sensor response approach according to the invention in a MIMO system with M transmitter elements and N receiver sensors and representing any waveform feature (as amplitude, delay, phase, frequency or polarisation),

Fig. 3 shows a received signal digitization scheme according to the invention in a MIMO system with M transmitter elements and N receiver sensors and representing any waveform feature (as amplitude, delay, phase, frequency or polarisation),

Fig. 4 shows a wireless communication system with time reversal technology,

Fig. 5 shows a single user time reversal MIMO sensor system with M transmitter emit- ters and N receiver sensors, Fig. 6 shows a time reversal approach according to the invention with M transmitter elements and three receiver sensors by mapping the data stream with an sensor pattern diagram (representing any waveform feature as amplitude, delay, phase, frequency or polarisation),

Figs. 7a-d show illustrations of wavefront feature coding across a sensor array with respect to amplitude, phase, delay or polarisation, respectively,

Figs. 8a-b show illustrations of block level overlay sub state coding/modulation using binary phase shift keying (BPSK) for base communication and 8PSK for overlay coding,

Fig. 9 shows an overlay on single data stream base data,

Fig. 10 shows an overlay on orthogonal space multiplex base data, and

Fig. 11 shows an overlay on non-orthogonal space multiplex base data using sensor 'pin pointing' with focused physical channels.

Best Modes for Carrying out the Invention

The basic idea of the approach according to the invention is the realization of sensor footprint identification as a part of the data stream in order to provide a physically (spatially coded) transmission for secure communication. Additionally, the footprint can be used to help save the radio spectrum and improve the channel capacity. Instead of emitting the whole data stream, the Tx only sends a part of the data stream to the Rx side. The other part of the data stream is used to identify the sensor footprint by applying a mapping or overlaying table. This mapping or overlaying table is also known at the Rx side. Therefore, from the specific individual sensor footprint, the mapping part of the transmitted data stream can be estimated by the Rx, along with the received data, the Rx is able to acquire the whole data sequence with less spectrum occupation and/or higher capacity achievement.

The most promising aspect of the information is to use the waveform coding for secure communication, so that the information that is to be sent from the Tx to the Rx is sent as a base part, whereas an overlaid part is used either for transmitting additional se- cure information and/or for conveying a secret key shared only by the Rx and the Tx. The secret key is thus used for encoding and decoding of the base part (this idea is similar to the invention described in US provisional application no. 60/716,945 by the same applicant). An eavesdropper cannot decode the information, since the spatial re- sponse at the eavesdropper is different from the spatial response at the Rx due to different positions of the sensor arrays of the Rx and eavesdropper, respectively.

The following examples are concentrated on overlaying by use of the amplitude of the wavefront at the receiver sensors as shown in Fig. 7a. However, it is also possible to use other waveform features such as phase (as shown in Fig. 7b), delay (as shown in Fig. 7c), polarisation (as shown in Fig. 7d), frequency, or combinations thereof.

In the case of radio communications, the properties of the wavefront are sensed by radio transceivers with antennas. The same techniques can be applied to other types of waves, such as acoustic waves by use for instance of microphones, by sensing a subset of the above properties.

The overlay information can be coded in a higher order version of the base communication modulation scheme, such as digital phase, amplitude or frequency modulation, e.g. ASK, QAM, PSK, or FSK. If for instance the base communication is performed with binary phase shift keying (BPSK), higher order phase shift keying states can be used for the overlaying. This is possible in a low rate overlay block coding, where the higher (via averaging gain) block SNR allows for finer granularity of detection/states than the one possible in the base communication (see also Fig. 8a and 8b). If the base commu- nication for example is binary frequency shift keying (BFSK) modulated, higher order FSK overlay states can be used with fractional frequency shifts (of the frequency shift representing the difference of base communication '0' and 'V states). This overlay frequency shift (at block level) can for example be extracted from the frequency synchronisation loops of the receiver.

Three distinctions are made with respect to overlaying:

1. Number of data streams on the base communication

We distinguish between two cases: single data stream communication on the base communication versus (space) multiplexed MIMO communication (orthogonal or non- orthogonal): Overlaying on single data stream communication means that the simplest single receiver sensor controlled overlay coding is most obvious, as control over more sensor states at the same time will require an effort of establishing transmitter weights that most likely should have been used to establish space multiplexing in first instance.

An overlay on multiple (space multiplexed) data streams is advantageous. Decoding of the individual base data streams at the receive side inherently provides the sensor isolation needed for best possible overlay coding. The overlay communications can be in- dependency controlled. The easiest way to visualize this is by assuming that each data stream is associated with a single sensor.

2. Physical versus virtual or transformed waveform coding

Physical waveform coding relates to the receiver sensors sensing the actual physical state of the media waveform. This is the most straight forward approach. However, if an orthogonalization transformation is used (like SVD), virtual orthogonal channels (transformed orthogonal channels) are launched. Each of these virtual channels can be interpreted and exploited as a 'virtual waveform', which is sampled by the virtual sensors (extracting of the virtual channels at the receiver). All of the overlay coding schemes can be applied in the virtual domain (the overlay coding does not have to be orthogonal). For example, in the low rate block coded cases some offsets can be applied between virtual channels so that the receiver synchronisation loops can extract the overlay decoding after decoding the base communication. The advantage is that overlaying is straight forward for orthogonal multiplexing schemes.

3. Block vs. symbol level overlaying

If the base data is transmitted at the symbol level using a conventional modem and if the overlay data is conveyed at a block level, where the base data detection/constellation diagram and channel states are close to static over the block trans- mission (several symbols durations), the response over the receiver sensor array averaged over the block of symbols will bear the overlay data information. This can be done without affecting the base communication since the overlay can be designed independently or transparently.

An overlay at the base data symbol level will affect the base communication and must thus be designed jointly with the base communication. The expected complexity is much higher in this case. In a MIMO system the optimum capacity can be achieved by orthogonal channel (SVD) decomposition, assuming infinite modem flexibility. Common digital modulation forms are inherently discrete and result in very rough throughput (capacity) increments for lower order modulations (high noise applications). Overlay waveform coding provides throughput granularity at lower SNR and brings the total throughput closer to the theoretical maximum. Alternatively to waveform coding, other types of coding could be used (e.g. turbo codes) to approach the capacity limit.

Consequently, the overlaying itself has three main features. Firstly, the communication information interpretation is unique to a transmitter element array and receiver array pair. Thereby spatial security with respect to information 'leakage' to unsolicited terminals/nodes is provided. Secondly, the overlaying can be performed on top of existing base communication schemes and can, in the low data rate situation (e.g. for key/handshaking information conveyed at block level only etc), be done nearly invisibly to any eavesdropper and without disturbing the base communication. As an example, the base information can carry bulk information/data while the overlaid part conveys the interpretation keys for the base/bulk data. Thirdly, in a joint optimised base and overlaid transmission/interpretation scheme, the overlaid part provides more degrees of freedom to approach higher capacity or throughput.

We distinguish the following cases:

a) Overlaying on single data stream communication

If there is only one data stream being transmitted as the base communication, an em- bodiment of the invention uses a time reversal transmission to produce amplitude modulated waveform overlaying on the sensor array, such as shown in Fig. 9. This is further explained below by example.

The Time Reversal (TR) scheme is just one very simple (and not the most effective) method to do overlaying, since it is simple to create and simple to decode. The TR scheme is chosen as the later example, since it can code information response at the receiver side with just one controlled sensor, and the rest of the sensors just being subjected to random/array or spatial noise. However, a person skilled in the art will recognise that other and more efficient schemes can be utilised.

Higher order modulation schemes, such as phase modulation, will require some sort of reference, to work properly. This reference could be provided from the base data flow or through a preamble part requiring constraints in the communication protocol/data format. However, if the overlaid data flow should be conveyed as transparently to the base data or communication protocol as possible, differential schemes should be used, meaning at least two sensor responses need to be fully controlled simultaneously. Simple time reversal schemes cannot facilitate this, and more complex transmission algorithms need to be employed in this case.

This motivates looking into multiple data stream transmission.

b) Overlaying on multi data stream communication (space multiplex)

The methods and embodiments can also be expanded to be used in a multi datastream MIMO system. Space multiplexing schemes automatically provide full control of each data stream, in the case of orthogonal transmission, where a virtual wave front is cre- ated (See Fig 10). In a non-orthogonal case, sensor 'pin-pointing' can be performed (to provide one data stream per receive sensor, see Fig. 11). Thus all data streams are associated with either a virtual channel sample (in the transformed eigen domain) or directly in a physical waveform.

For non-orthogonal space multiplexing (as for instance described in US provisional patent application No. 60/719,984 by the same applicant) the transmitter weights seek to maximize the receiver sensor isolation (as each data stream is associated with a single receiver sensor). Consequently, this scheme is also an implementation of simultaneous and individual receiver sensor state (waveform) control, allowing best possible overlay coding.

Simple examples of implementation

Based on the above mentioned main idea, a Sensor Response technique is proposed, which is the general overview of an enabling technique for the realization of this idea. In order to improve the stability, a second approach of the Digitization processing technique is applied at the Rx side. In addition, the use of a Time Reversal (TR) approach is simple (though not very effective) as it gives rise to better inherent spatial selectivity performance and improves the link reliability. Together with the capacity improvement achieved by the fewer data transmissions, the TR method describes the specific reali- zation of the idea behind the invention and achieves the ultimate purpose. As a consequence, three approaches are proposed in the following, namely the Sensor Response, digitization and Time Reversal approach, which will be presented in details in the following.

Sensor Response Approach In the sensor response approach, the data stream x (Length L) is split into two parts: One part B (Length b) is defined as the mapping part and is reserved to select the sensor pattern by which the following data is to be emitted. The other part D (Length d) is called the data part and is transmitted to the intended Rx, which is named as the data part. Only the data part is propagated through the channel and the mapping part is used for identifying the sensor pattern by which the next data part is to be transmitted using conventional any modem technique. Thus, x = {B, D}, where B is the mapping part and D is the data part.

In the following, the data part D will often be referred to as the base part and the map- ping part B as the overlaid part or waveform coded part. Preferably the base part carries the information, which is to be sent from the Tx to the Rx, while the overlaid part carries the shared key or footprint used for secure communication, so that a third party cannot eavesdrop on the communication.

The sensor pattern at the Rx side is identified by the signal feature (amplitude, phase etc) strength/focus across all the Rx sensor arrays. The different possible patterns correspond to different bit combinations in the overlaid data B. The intended Rx is assumed to have the full knowledge of the mapping table according to which the bits in B are mapped into sensor patterns. From the sensor pattern, the Rx can recover the en- tire original data stream: the mapping part can be obtained from the demapping table of the perceived array pattern and the data part is acquired from the decoding process of the received signal. As a consequence, the radio spectrum is efficiently utilized. In addition, since both the Tx and intended Rx know the specific sensor pattern, but the specific sensor pattern cannot be successfully sensed by eavesdroppers, the link secu- rity is enhanced with low probability of interception (LPI).

In this Sensor Response approach (see Figure 2), any TX array can create infinitely many sensor patterns. Increasing the number of patterns increases the size of the mapping table and decreases the discernibility of the patterns (this depends on the number of Tx elements). The sensor pattern can be established randomly; hence the number of sensor patterns the Tx may create is infinitive, which leads to the increasing volume of the mapping table. However, this infinity value is limited by the number of Tx emitters M, which decides the signal quality. The minimum requirement is M > 1 to establish the various sensor patterns. Moreover, the mapping table at the Rx side is required to be dynamically updated in order to match the possible sensor pattern the Tx established.

The advantage of the sensor response approach is that an infinite number of sensor patterns is possible, while the disadvantage is that the mapping table of the sensor patterns at the Rx side needs to be dynamically updated. However, if the mapped informa- tion is continuous or analogue, then the sensor approach is straight forward and an obvious approach to take (for example analogue voice could be overlaid/modulated via offset amplitude or phase onto base data that could be digital).

Received Signal Digitization Approach With the aforementioned Sensor Response technique, the correspondence between the sensor pattern and the data is established. However, the received signal is not stable due to the presence of the interference and noise. In order to improve the signal stability and discemibility of the patterns, the received signal can be digitized, which extends the Sensor Response approach. The principle of this processing technique is de- fining a digitization threshold on the signal amplitude and/ or phase on the Rx sensor arrays in order to determine the sensor pattern (see Figure 3). At the Rx side, the signal across the Rx sensor arrays is sampled. At each sample position (sensor response and/or base data substream), the received signal is quantized with a certain number of binary digits. The total bits consist of a bit sequence which can be used to identify the sensor pattern.

In Fig. 3 a first wavefront 10 and a second wavefront 11 are illustrated. When the strength/level/focus of the wavefront feature (e.g. amplitude, phase etc.) is equal to or above the digitization threshold T (with reference numeral 20), the receiver side inter- prets the signal as T. If the wavefront features level is below the digitization threshold T, the receiver side interprets the signal as ¹O'. Thus, the signal for the first wavefront 10 is interpreted as '0' at Rx₁ and as 'V at Rx_m, while the signal for the second wavefront 11 is interpreted as '1' at Rxi and as '0' at Rx_m. For simplicity, only 1 bit is used as an example. If the received signal parameter strength/level (amplitude, delay or phase etc.) is above the threshold, it is noted as symbol T; otherwise, it is ¹O'.

_ J 1 Signal Strength of Intended Rx Antenna ^ Threshold ^{y ~} \ 0 Signal Strength of Intended Rx Antenna < Threshold Compared to the previous Sensor Response approach, this processing technique provides stability of interpreting the level of received signal feature/parameter. However, the number of the sensor patterns (receive wavefronts) is bounded by the number of digital bits (I_digi_tai) used in the digitization step and the number of the samples/ sensors (N_sampn_ng)- The following equation is used to calculate the number of the sensor pat- terns.

N_sψM = 2''^ (2)

In the example of 1 bit digitization process, the number of the symbol N_symb_Oι is 1 , namely 0 and 1. If N_sarrφn_ng = 4, the number of the sensor patterns can be identified by the intended Rx is 2^1"4 = 16. Additionally more bits can be applied during the digitization procedure. For example, if 4 decision regions are used, then the digitization gives 2 bits, i.e. 4 symbols are established, namely 00,01,10,11. Hence, the signal parameter strength across the Rx sensor arrays is split by 4 regions, with each symbol identifying one region. If N_sampι_m- _g = 4, the number of possible combinations of these 4 regions is 2^4'4 = 65536. Thus, the possible number of the sensor pattern increases considerably. The signals can be digitized based on the signal amplitude, polarization, delay or frequency or any combination thereof.

This method enhances the received signal stability by applying the digitization process. However, the processing technique is still difficult to manipulate and the single sensor spatial focusing behaviour is still not explicit because of the induced interference to other sensors.

In order to improve the spatial selectivity performance, the Time Reversal technique applied in the MIMO system is proposed as the further extension of these two methods.

The technique is derived from the pattern diversity technique, adding a digitization threshold at the Rx side. The advantage of the approach is that the received signal is more stable than the simple beam forming technique. The disadvantage is that the ap- proach only has a finite number of sensor patterns and is limited by the number of digital bits {/_digi_tal) used in the digitization step and the number of the sensors {N_sgmp,ι_ng). Also, the manipulation is still difficult to realize.

Time Reversal Approach on single data stream overlaying

Based on the principle of the combination of the data with the sensor pattern establishment, the sub-optimal non-orthogonal parallel channel is utilized by employing the Time Reversal (TR) technique in order to create the sensor pattern and improve the signal quality, since the TR approach has the potential performance of spatial selectiv- ity, which can be employed for the sensor pattern establishment. In addition, the numerical complexity in the TR approach is simpler compared to the orthogonal parallel channel realized by the SVD technique. This leaves less processing burden at both the Tx and Rx sides.

The Time Reversal method applied in the MlMO system results in non-orthogonal transmission and it is anticipated to attain the link reliability enhancement directly and realize the capacity improvement indirectly with lower complexity requirement. In addition, the Time Reversal technology employed in the multiple element system can lead to three merits of the performances: time compression, spatial focusing and diversity gain.

Orthogonal space multiplex overlaying

The orthogonal space multiplexing technique is achieved by using the SVD technique (see Fig. 10), which decomposes the channel as: H = U-Z-V*

,

where H matrix holds the channel transfer functions. The unitary matrices V and U contain the right (input) and left (output) singular vectors of H respectively (used to 'expose' and 'extract' the virtual orthogonal channels), i, is the M dimensional identity matrix, Σ is a diagonal matrix that contains the singular values X₁ of H (i.e. the gains of the virtual channels. The transformed domain orthogonal channels are automatically aligned with respect to phase using typical SVD algorithms, so any overlay phase coding is straight forward to implement. The disadvantage of orthogonal space multiplexing using SVD is that the technique needs both receiver and transmitter weights (see Fig. 10), but overlaying can be made transparent to this if low rate block level coding is used (the waveform coding corresponds to the whole constellation diagram of the base data being shifted from one sub- stream to the next (i.e. the sensors). Typical receiver synchronization loops for phase, frequency etc. can then be used to extract the overlay information via comparison of these synchronization loop responses (offsets).

Non-orthogonal space multiplex overlaying Non-orthogonal space multiplexing can be achieved for instance by using a form of optimum combing [3][4] that to a certain degree isolates signal stream to different receiver sensors. The optimum combining algorithm has been developed for uplink (receive) situations but the same weights can be used as suboptimum downlink weights (creating focus to particular sensor element) (see also US provisional application No. 60/791,984). The transmitter filter weights are then selected as R-¹^ H-^* (in order to maximize the uplink signal to interference plus noise ratio (SINR) for the /c-th data stream), where the matrix R_NH is the spatial covariance matrix of the noise plus interference from the other data streams R_N+7 =R^R_{7 1} R₇ = H^₁ Hj₁ , and (-)^*,(-)^r indicate the conjugate and transpose of the argument (^■), respectively. R_W is a diagonal matrix with diagonal elements equal to the noise variance N [5].

The effect of this space multiplexing can be interpreted as sensor 'pin pointing' with the advantage that only transmitter weights are needed to establish spatial substreams, while still providing independent simultaneous control of all sensors (see Fig. 11).

Time Reversal Principle

In principle, the TR operation is a two stage process [6]:

In the first step, the transmitter estimates the channel IR h(t) of the intended receiver. This leads to the filter weight g(t) at the transmitter side to being equivalent to the complex conjugated time-reversed channel IR h(t), namely g(t) = h*(-t). The channel is assumed to be stationary, which means the estimated channel IR is equivalent to the channel IR of the intended user where the transmitted signal will be propagated. In the second step, the actual transmitted signal is acquired by convolving the input data stream z(t) and the TR filter weight g(t), that is: s(t) = z{i) © g{t) = z(t) © h*{-t) (4)

where symbol ® represents the convolution operation and * expresses the complex conjugate.

At the receiver end, the signal y(t) is obtained as the convolution between the transmitted signal s(t) and the channel IR h(t) (see equation (5)):

y{t) = s(t) Q k{i) = (z(ή ® h*(-tfj φ h(t) = z(t) <8> (h*{-t) «g> h(tγj = z{t) 0 R^t) (5)

where R_hh(t) denotes the autocorrelation function of h(t). Consequently, the received signal is the convolution of s(t) with the autocorrelation function of the channel IR, which holds the property that at t = 0, the maximum value of R_hh(t) is achieved and equals to the power of the input signal.

However, this benefit of the autocorrelation function appears to be predominant, when the TR is applied with several links, for instance, the multiple element system. This is because the suppression of the side lobes may not be significant in the single link transmission. In the multiple element system, the received signal is the summation of the autocorrelation function in each link. After the summation of autocorrelation func- tions, the suppression of the correlation side lobes is achieved, due to the coherent addition of the main peaks and the non-coherent summation of the side lobes.

MIMO Time Reversal Operation

The application of the TR in the MIMO system refers to employing the TR processing technique at each Tx element. Figure 5 illustrates a single user TR MIMO system with M transmitters and N receivers. The TR processing in this diagram is described in the following.

Filter Weight Achievement The input signal z((t)) is transmitted through M Tx elements. The filter weight g_m(t) denotes the complex-conjugated, time-reversed channel IR h^* _m (-t) obtained at the m^th

Txelement. Since the Rx has N sensors, the Tx can obtain the channel IR to all these Rxsensors. When communicating to receiver sensor n, each transmitter m will apply a filter that is given by s.,M=C(-0- ⁽⁶⁾

Transmitted Signal Establishment

The transmitted signal intended for the n^th receiver sensor sm,n(t) is established as

^Sm ,n (0 = * (0 ® g ,n ,n (t ) (7)

Received Signal ym,n(t) is the signal received from the m^th Tx emitter at the n^th intended Rx sensor. In the absence of noise, it is the convolution of the transmitted signal sm,n(t) and the channel impulse response (IR) h_mn(t) from the m^th Tx emitter to the n^m Rx sensor.

³(0 ® *^Ω"^to (0 ⁽⁸⁾

where R^aυt0(t) is the autocorrelation function of the intended channel IR.

ym,n'(t) is the signal received at other non-intended sensors Rx_n-

2(0 ® Λ"" (0 ⁽⁹) where R^cross(t) is the cross-correlation function between the intended and other channel IRs. This is defined as the leakage signal (interference) on other Rx sensors.

The total received signal at the intended receiver sensor is the coherent sum of all autocorrelation functions given in equation (8), while the leakage signal at other sensors is the non-coherent addition of all cross-correlation functions in equation (9).

Due to the non-zero leakage, communication to different receiver sensors will be non- orthogonal. Ideally the signal on the intended receiver should be non-zero, and the signal on the non-intended receivers should be zero. However leakage exists. The leakage signal can be interpreted as 'space noise' and sets limits for individual sensor state (wavefront) control. Sensor Pattern Establishment by TR Approach

As previously mentioned, the TR approach has the benefits of spatial selectivity, which can be used to establish the sensor pattern. The principle is to use the mapping part of the data to identify the corresponding sensor pattern at the Tx side and to focus on the intended Rx sensors based on the selected pattern. At the Rx side, the distribution of the signal strength on the Rx sensor arrays is utilized to identify the sensor pattern. Unlike the Sensor Response and Digitization methods, the signal strength in the Time Reversal approach only refers to the signal amplitude, since the amplitude is more sta- ble than the phase. Therefore, the distribution of signal amplitude on the Rx sensor arrays determines the sensor pattern.

Transmitter Side

At the Tx side, it is assumed that a data sequence is the original data stream with length L. Instead of emitting the whole data sequence into the propagation channel, the first b bits are preserved to map with the possible sensor patterns and determine which kind of the sensor pattern will be utilized, while only the next d bits are propagated through the channel. Similar behaviour will be iterated for the whole data stream. According to the mapping part of the original data, the sensor pattern is connected with this part of the data stream and predefined at the Tx side. Applying the Time Reversal method to establish the different sensor pattern shape based on the mapping part of the original data, the distribution of the signal strength is defined. For simplicity, it is assumed that the sensor signals are quantized to 1 bit. If absolute threshold levels are used to determine the bit values at each sensor, then all 2^b sensor patterns are possi- ble. If relative threshold are used, the all 1's and all O's are ambiguous. Then the number of total sensor patterns is: N_A = 2^b - 2. (relative thresholds can be used to define 1 's at more than one sensor).

In this approach, it is assumed that the Tx emitters can focus on not only the single sensor, but also the multiple Rx sensors, i.e, double, triple, etc. Thus, a method that uses the binary symbol T and O' to discriminate the intended Rx sensor is proposed. The combination of the binary data '0' and '1 ' is used to define the different sensor patterns. 'V indicates the corresponding Rx sensor is the intended one, while '0' expresses that Rx sensor is not intended by the Tx. _Ώ . . _. . _ , f 1 Intended Rx Antenna Received Sig ^{6^}nal Level = s I_{1 n} 0 ^ Ts₇ on-i •n _*tend _Jed J - R_ox A _Λ n _*tenna

In order to minimize the hardware complexity, the number of the Rx sensors is assumed to be equivalent to the mapping part of the original data stream b bits, which decides the number of the sensor patterns NA, in order to reduce the number of the Rx sensors and leave less complexity at the Rx side,

N = b (10)

N_A = 2^fe - 2 = 2^jV - 2 (11)

Receiver Side

At the Rx end, the receiver detects the signal strength of the Rx sensor arrays. The sensor pattern is distinguished by the distribution of the signal strength on the Rx sen- sor arrays. In addition, the Rx has the prior knowledge of the mapping table. From this table, the Rx knows the mapping part of the original data by demapping the sensor pattern. In addition, the data part of the original signal is directly received by the Rx. These two parts consist of the total signal length. Thus, if the Rx obtains the first transmitted bit 1 from the sensor pattern 2, it knows the first four bits original data is 0011. There- fore, the whole original data stream is understood by the Rx from the emitted data and its perceived sensor pattern.

Pattern mapping

In the aim of clearly illuminating the proposed approach utilized by the TR technique, one example of utilizing 3 bits to identify the sensor pattern is explained in this section (see Figure 6). The original data stream has the form of x = (00111010). At the beginning, the first three bits '001' are taken as the mapping part. The mapping table shown in table 2 is used. Clearly, it has been derived using relative thresholding (at least a single 1 appears in any 3bit combination). From Table 2, it can be seen that '001 ' matches sensor pattern 2. Therefore, Rx1 is assumed as the intended receiver and the data part T is emitted to focus on Rx1. At the receiver side, the receiver detects that Rx1 has a higher signal strength. Thus, the receiver knows that the sensor pattern 2 is utilized, which means that the mapping data is '001 '. In addition, data part '1' is received from Rx1. Thus, the Rx reconstructs the data sequence and '0011' is decoded at the Rx side. For the next data sequence '1010', the data part '0' is transmitted by the sensor pattern 5 which is dependent on the mapping part '101 ' from the mapping table. This mapping part '101' determines the Rx1 and Rx3 in the intended Rx sensors. The Rx detects this sensor pattern and recovers the original mapping part, together with the received data part. Finally, the Rx acquires the whole data stream transmitted from the Tx.

Table 2: Data Bits and Antenna Pattern Mapping and Demappiαg Table

Only a finite number of sensor patterns is achieved, limited by the number of different combination of the mapping data length.

As previously mentioned, the sensor patterns can be used for increasing the communication throughput/capacity or it can be used as overlaid information in order to achieve secure information between the Rx and the Tx.

In this document, the number of the sensor patterns is only taking the specific patterns established by the Tx emitter arrays into account. However, the difference among multiple sensor patterns can also be applied as the transmission signature. In addition, the signal detection employed in this paper is the received signals across the Rx sensor arrays. Another possible detection may be applied by making the discrimination from the time delay, frequency response, polarization etc.

The invention proposes a new approach to map the sensor pattern to a part of the original data stream. Since only a part of the data is transmitted using conventional modem techniques, the radio spectrum is saved and the capacity is improved. In addition, since only the intended Rx knows the sensor pattern established by the Tx ele- ments, the link security is enhanced and low probability of intercept (LPI) is possible.

The invention has been described with reference to a preferred embodiment. However, the scope of the invention is not limited to the illustrated embodiment, and alterations and modifications can be carried out without deviating from said scope of the invention. References

[1] http://www.atis.org/tg2k/_channel_capacity.html 05.06.2005

[2] Zushun Song, "Modern Communication Theory," Publishing House of Electronics Industry, Beijing, China. ISBN: 7-5053-6255-0, pp. 210, 2001.

[3] J. Winters, Optimum combining in digital mobile radio with cochannel interference," in IEEE journal on Selected Areas in Communications, Special Issue on Mobile Radio Communications, vol. 2, No. 4, pp. 528-539, July 1984.

[4] J. Winters, "Optimum combining for indoor radio systems with multiple users," in IEEE Trans. On Communications, vol. 35, No. 11 , pp. 1222-1230, Nov. 1987.

[5] R. Vaughan and J. B. Andersen, "Channels, propagation and antennas for mobile communications," IEEE Press, UK, pp. 632, 2003.

[6] Xin Zhou, Jimena Martinez Llorente, Axel Adenet.Christophe Lemasson, Patrick Eggers, "Assessment of MISO Time Reversal for short-range communica- tions in the 5GHz ISM band" to be submitted to IEEE transactions on Antenna and Propagation

[7] G.J. Foschini and M.J. Gans, "On limits of wireless communications in a fading environment when using multiple antennas," in Wireless Pers. Commu., vol.6, pp. 311-335, Mar. 1998.

[8] I.E. Telatar, "Capacity of mulit-antenna Gaussian channels," in Eur, Trans. Telecommun., vol.10, No.6, pp. 585-595, Nov./ Dec. 1999.

[9] J. B. Andersen, "Array gain and capacity for known random channels with multiple element arrays at both ends," in IEEE J. Select Areas Commun., vol. 18, pp. 2172-2178, Nov. 2000.

Claims

1. A method for sending information in a communication system comprising a transmitter side and a receiver side, said receiver side having a number of sensors and said transmitter having a number of emitters, wherein a) the transmitter side transmits a data stream containing information, which is divided into a first part and a remaining part, in such a way that a certain spatial sensor response is produced at said number of sensors on the receiver side, said sensor response carrying at least the first part of said information contained in the data stream, b) the receiver side detects and interprets said spatial sensor response and extracts at least the first part of said information, and c) the receiver side detects and interprets the remaining part of the information.

2. A method according to claim 1 , wherein the sensor response is interpreted with respect to signal characteristics, such as amplitude, phase, polarization, delay or frequency, of the signals detected at the receiver sensors.

3. A method according to claim 1 or 2, wherein the sensor response comprises physical waveform samples across a sensor array, retrieved at specific sensor loca- tions at the receiver side.

4. A method according to claim 1 or 2, wherein the sensor response comprises transformed waveform samples in a transformed eigen domain, whereby each created channel or eigen mode is acting as a virtual sensor.

5. A method according to any of the preceding claims, wherein the first part of the information is used for secure communication.

6. A method according to any of the preceding claims, wherein at least the first part of said information is coded using differential sensor responses between two or more individual sensors of said number of receiver sensors.

7. A method according to any of the preceding claims, wherein the data stream at the transmitter side is divided into a base part and an overlaid part, and the receiver side decodes the base part using standard decoding methods and decodes the overlaid part using the sensor response detected.

8. A method according to claim 7, wherein the receiver side uses interpretation information, such as an overlay coding table or a modulation transfer function, so as to extract the overlaid part of the information based on the sensor response received at the receiver side.

9. A method according to claim 8, wherein the interpretation information at the receiver side is dynamically updated.

10. A method according to any of the preceding claims, wherein the sensor response at the receiver side is digitised with respect to signal characteristics, such as any combination of amplitude, phase, polarization, delay or frequency.

11. A method according to claim 10, wherein the sensor response at the receiver side is digitized using a number of detection regions.

12. A method according to claim 11, wherein each of the received signals at said number of sensors at the receiver side is compared to said number of dectection regions.

13. A method according to any of claims 10-12, wherein digitisation conditions are dynamically exchanged between the transmitter side and the receiver side.

14. A method according to claim 7 or claim 7 and any of claims 8-13, wherein a time reversal transmission scheme is used for the transmission of the base data and produces amplitude modulated waveform overlaying on the sensor array.

15. A method according to any of claims 7-14, wherein non-orthogonal parallel channels are used for sending the data streams from the transmitter side to the receiver side and for separating the overlaid data on the sensor response.

16. A method according to claim 14, wherein the transmitter side prior to step a) estimates an impulse response h(t) for a channel between the transmitter side and a given receiver sensor and subsequently generates a filter weight for the channel, said filter weight being the time reversed conjugate h*(-t) of the estimated impulse response, and wherein the transmitter side subsequently convolves the data stream with the filter weight and in step a) transmits the resulting convolved signal to the receiver.

17. A method according to any of the preceding claims, wherein a time reversal trans- mission scheme is used for coding the transmitted data stream to produce a pulse/time modulated waveform, such as a pulse-position modulated (PPM) waveform, overlaying the sensor array.

18. A method according to any of the preceding claims, wherein a polarisation state transmission scheme is used for coding the transmitted data stream to produce polarization modulated waveform overlaying the sensor array.

19. A method according to any of the preceding claims, wherein overlay information is coded in a higher order version of the base communication modulation scheme, such as digital phase, amplitude or frequency modulation, such as ASK, QAM, PSK, or FSK.

20. A communication system for sending information, said system comprising a transmitter side and a receiver side, wherein

- said transmitter side comprises transmitter means, such as a number of transmitter elements, for transmitting a data stream containing information and being adapted to transmit the data stream with a spatial pattern, which is at least dependent on a first part of said information, and

- said receiver side comprises a number of sensors, which are adapted to detect a spatial sensor response, said receiver side being adapted to convert the spatial sensor response to information, so as to extract at least said first part of the information from said data stream.

22. A communication system according to claim 21 , wherein the transmitter means comprises two or more transmitter antennas, and the receiver sensors are receiver an- tennas.

23. A communication system according to claim 21 , wherein the transmitter means comprises two or more acoustic transducers and the receiver sensors are acoustic receivers.

23. A communication system according to any of claims 20-23, wherein the transmitter side is adapted to dynamically update transmission conditions in order to produce the desired spatial sensor response on the receiver side based on said first of the information.

24. A communication system according to any of claims 20-23, wherein the transmitter side is adapted to split the data stream up into a base part and an overlaid part.

25. A communication system according to any of claims 20-24, wherein the receiver side is adapted to interpret the detected sensor response using an overlaying table.

26. A communication according to claim 25, wherein the receiver side additionally comprises decoding means known per se for extracting the base part of the data stream.

27. A communication system according to claim 21 or claim 21 and any of claims 23- 26, wherein the system is a multiple input multiple output (MIMO) system, said transmitter side having a plurality of transmitter elements, and said receiver side having a plurality of receiver sensors, a number of overlay non-orthogonal channels being gen- erated between said transmitters and receiver sensors.

28. A communication system according to claim 21 or claim 21 and any of claims 23- 26, wherein the system is a multiple input multiple output (MIMO) system, said transmitter side having a plurality of transmitter elements, and said receiver side having a plurality of receiver sensors, a number of orthogonal channels being generated between said transmitters and receiver sensors for base communication and a number of other orthogonal channels being dedicated to overlaying.

29. A communication system according to claim 21 or claim 21 and any of claims 23- 28, wherein the transmitter side additionally comprises filter weighting means for weighting the data streams transmitted from the transmitter element.

30. A communication system according to claim 29, wherein the weighting means are adapted to weight the data streams according to a time reversal scheme.