WO2008004156A1

WO2008004156A1 - Blind amplitude estimation for received symbols

Info

Publication number: WO2008004156A1
Application number: PCT/IB2007/052426
Authority: WO
Inventors: Alexander Lampe; Stefan Mueller-Weinfurtner
Original assignee: Nxp B.V.
Priority date: 2006-07-03
Filing date: 2007-06-22
Publication date: 2008-01-10
Also published as: EP1876786A1; US8290090B2; US20090181718A1; EP2039104A1

Abstract

The present invention relates to a receiver apparatus and method of blind amplitude estimation for a received complex signal, wherein the complex signal is analyzed to determine first and second order statistics of real part and imaginary part of the complex signal. Based on a predetermined non-linear function, amplitude scaling information is pre- calculated and used to derive a desired amplitude scaling information for said determined first and second order statistics. Thereby, a low complexity method for blind amplitude estimation for unknown data symbols distorted by an unknown amount of noise can be provided.

Description

BLIND AMPLITUDE ESTIMATION FOR RECEIVED SYMBOLS

The present invention relates to a receiver apparatus for receiving a complex signal via a digital transmission channel, and to a method of blind amplitude estimation for a received complex signal. In particular, the present invention relates to a communication system in which an M-ary modulation scheme is employed for digital transmission.

In a digital communication system, digital symbols representing information are transmitted between different devices, e.g., between a base station device and a mobile terminal, to exchange information.

At the receiver side, the received signal is processed to obtain a sequence or stream of digital samples and these samples may be represented as complex symbols. For example, the received signal may be filtered, amplified, and mixed down to base band using in-phase and quadrature local oscillators, and after analog-to-digital (AJO) conversion and synchronization processing, a stream of complex received symbols is obtained. Each symbol in the complex symbol stream may be represented as a sum of a real part and an imaginary part.

Newer 3rd generation (3G) cellular communication systems employ wideband code division multiple access (WCDMA) technology. An extension of the WCDMA standard is known as high speed downlink packet access (HSDPA) and has recently been standardized within the 3rd Generation Partnership Project (3 GPP) standardization organization. HSDPA introduces known technologies such as higher order modulation and incremental redundancy to the 3GPP Universal Mobile Telecommunications System (UMTS) standard. The higher order modulation introduced in HSDPA is M-QAM, and more particularly 16-state Quadrature Amplitude Modulation (16-QAM).

Fig. 2 shows a schematic complex diagram, wherein the horizontal or x-axis corresponds to the in-phase (I) channel (i.e. real part of the complex symbol) and the vertical axis or y-axis corresponds to the quadrature (Q) channel (or imaginary part of the complex symbols). Generally speaking, 16-QAM is achieved by modulating two four-level pulse amplitude modulated (PAM) signals onto two respective orthogonal carriers (I and Q), providing 4² = 16 possible symbol representations. Accordingly, a 16-QAM symbol includes phase information based on the respective I or Q orthogonal carrier and amplitude information, whereas Quadrature Phase Shift Keying (QPSK) detection, by comparison, includes only phase information.

In Fig. 2, every black area or dot represents one complex symbol with its phase and amplitude information. The amplitude information corresponds to the radial distance of the dot from the origin of the coordinate system and the phase information corresponds to the angle with respect to the positive x-axis. Due to the fact that 16 symbols can be represented in the 16-QAM system, each symbol corresponds to a 4-bit word comprising four bits So to S3. The binary representations of the complex symbols are indicated above the dots in Fig. 2.

Receivers in WCDMA systems rely on a reference signal, such as a time- multiplexed pilot symbol or code-multiplexed pilot channel, to calculate estimates of a radio channel's response. Typically, the channel gain and phase of the Common Pilot Channel (CPICH) is estimated once per slot for this purpose. The 16-QAM data, which is transferred on the High Speed Physical Downlink Shared Channel (HS-PDSCH), requires processing of phase and amplitude information to recover the information in the data. While the phase information can be reused from the CPICH based measurements, the gain offset between the CPICH and HS-PDSCH, which is unknown to the receiver, must be determined separately in order to establish an amplitude reference. Thus, in order to correctly demodulate the transmitted symbols, the amplitude or, equivalent Iy, scaling factor of the transmitted symbols has to be known. In many systems, the amplitude is not known a-priori on the receiver side and therefore has to be estimated. The accuracy of this estimate has direct impact on the link level performance in terms of frame error rate and/or data throughput. If no information on the amplitude of the transmitted symbols (e.g. relative to known pilot symbols) is signaled, this amplitude has to be "blindly" estimated by the receiver. I.e., based only on the unknown received data symbols. In a real world transmission system, the received symbols are additionally distorted by noise and/or interference with unknown statistical parameters. However, conventional methods for blind amplitude estimation require quite complex calculations.

The US 2004/0096015 Al discloses estimation of the average value of the in- phase/quadrature component of a received signal as well as signal power by averaging the magnitudes of a sufficiently large number of received signal samples. These values are used by a hard decision unit to estimate the amplitude of the data symbols. However, a specific method of calculation of these amplitude estimates is not given or cited.

It is therefore an object of the present invention to provide a low complexity scheme for blind amplitude estimation for unknown data symbols which are distorted by an unknown amount of noise.

This object is achieved by a receiver apparatus as claimed in claim 1 and by a method of blind amplitude estimation as claimed in claim 6. Accordingly, an amplitude scaling information required for amplitude estimation is obtained by addressing a memory (e.g. look-up table) storing pre-calculated amplitude scaling information, based on first and second order statistics, which can be easily calculated. Thereby, the blind amplitude estimation is reduced to a calculation of simple first and second order statistics and a simple memory look-up operation. Complexity is thereby significantly reduced.

The first and second order statistics may relate to an expectation value of the magnitude and a variance of the complex signal. The expectation value and the variance can be readily calculated from real parts and imaginary parts of the received complex signal.

The first and second order statistics may be obtained by taking a predetermined number of received symbols of the complex signal into account.

Furthermore, the determination means may be adapted to address the memory means based on a ratio between the first and second order statistics. This ratio may be selected to define a direct non-linear functional relationship between the ratio and the amplitude scaling information, so that the amplitude scaling information stored in the memory means merely reflects this non-linear functional relationship.

The predetermined non-linear function may be derived from probability density functions of noise and symbols provided in the complex signal. These probability density functions can be determined for the provided modulation scheme and related complex symbols to thereby pre-calculate the non-linear relationship, based on which the amplitude scaling information is stored.

The present invention will now be described based on a preferred embodiment with reference to the accompanying drawings, in which: Fig. 1 shows a schematic block diagram of a receiver apparatus according to the preferred embodiment;

Fig. 2 shows a complex diagram indicating possible symbols of a 16-QAM modulation scheme; Fig. 3 shows a schematic flow diagram of the blind amplitude estimation method according to the preferred embodiment;

Fig. 4 shows a diagram indicating an amplitude distribution of complex symbols on the real or imaginary axis; and

Fig. 5 shows a diagram with a normalized non-linear function for obtaining amplitude scaling factors.

In the following, the preferred embodiment will be described in an initial general part and a later specific part for an example of a 16-QAM modulation based on a statistical signal model.

Fig. 1 shows a schematic block diagram of a receiver apparatus, as provided for example in a mobile phone or user equipment. The receiver device receives a signal via an antenna. The received signal may be a data chip sequence, which is a high-speed data series spread over a wide band, and which may have been modulated in a transmission apparatus by a multi-level quadrature amplitude modulation, such as 16-QAM or 64-QAM. The received sequence of data chips is supplied to a radio frequency (RF) unit 10 which may comprise a demodulator for demodulating the received signal, a despreader, a RAKE combiner or the like, so as to obtain a sequence of complex symbols at baseband frequency level. The baseband signal from the RF unit 10 is supplied to a detector unit 24 which performs a hard or soft decision for data symbols based on estimated amplitude scaling factors obtained from an estimation unit 26 which receives statistical information about the sequence of complex data symbols from a statistical analyzing unit 22. The detector unit 24 outputs hard or soft decision results, which are supplied to a decoder unit 28 in order to decode the received complex symbols and obtain the transmitted information.

In the following, the processing and functions at the statistical analyzing unit 22 and scaling factor estimation unit 26 are described in more detail.

Let y_k denote the received noisy complex valued symbol at time instance k , let K be the total number of received symbols taken into account, and f(x) be a function, which will be derived and described in more detail in the next section. E{ },V{ } denote estimates for the expectation value E{ } and the variance V{ }, respectively, which are calculated at the statistical analyzing unit 22. The operators abs(.),real(.),imag(.) denote the absolute value of the argument and its real or imaginary part, respectively.

Fig. 3 shows a schematic flow diagram of the basic procedure for estimating the unknown amplitude scaling factor a . In step SlOl an estimate for the expectation value of the magnitude of the received signal is calculated by the statistical analyzing unit 22, e.g., based on the following equation:

E{\y\} = ^-∑abs(real(y_k)) + abs(imag(y_k))

In a parallel step S 102, the statistical analyzing unit 22 calculates an estimate for the variance of the received signal (assuming zero mean), e.g., based on the following equation:

Of course, step S 102 may as well be performed after or before step SlOl. Finally, in step S 103, the scaling factor estimation unit 26 computes an estimate for the amplitude scaling factor a using a pre-computed function /(.) , e.g.:

The desired amplitude scaling factor a can now be applied by the detector unit 24 to the received sequence of symbols, thus normalizing the power of the symbol alphabet to unity.

The relationship expressed by the pre-computed function /(.) may be implemented by storing respective scaling factor values defined by the pre-computed function /(.) in a look-up table (LUT) 25 or other memory devices and addressing the LUT 25 by using the above statistical output values obtained in steps SlOl and S 102.

Thus, the received complex noisy signal is analyzed at the statistical analyzing unit 22 to determine first order (e.g. expectation value) and second order (e.g. variance) statistics of the real and imaginary part of the signal. These statistics are then used by the scaling factor estimation unit 26 to obtain an amplitude scaling factor from the LUT 25 which has been loaded based on signal model information. In particular, the statistics or any combination thereof is used as input or address for the LUT 25, which provides the desired amplitude scaling factor. Without loss of general validity the following specific example of the preferred embodiment relates to 16-QAM modulation. However, the above general procedure can be easily applied to any other (real or complex valued) symbol alphabet used in combination with linear modulation.

The baseband signal model introduced below is based on the assumption that the transmitted symbols, as supplied to the detector unit 24, are scaled by an unknown factor a and distorted by additive White Gaussian noise (AWGN):

with a : real valued amplitude scaling factor (to be estimated) x}^c) : complex valued transmit symbol at time index k

(16-QAM constellation assumed to have unity power) y[^c) : complex valued received symbol at time index k n[^c) : complex valued noise sample at time k, originating from a complex WGN process with total variance 2σ ² k : discrete time index, k e {0, 1, ... , K - 1}

It is noted that according to the above definition the noise term n_k ^(c) is also scaled by a . For the sake of simplicity, the complex variables are separated at the statistical analyzing unit 22 into their real and imaginary components and it is started with the following simplified model not yet including the factor a :

y_k = ^χk + ⁿk (²)

with x_k : real value of transmit symbol at time index k y_k : real value of received symbol at time index k n_k : real value of noise sample at time k, originating from a real valued WGN process with variance σ ²

The following assumptions are then made:

Real valued noise process with variance σ ² is White Gaussian and uncorrelated to signal, i.e.

E{x- n} = 0 (4)

Transmitted symbols are equally distributed (both in real and imaginary part) according to the probability density function (PDF)

^ΛW=ΪK-^O^)+8»(I-^()^)+8»(I+^O^)+8»(I+^O⁾)⁽⁵⁾

Fig. 4 shows a diagram indicating the distributions of real and imaginary parts in the concerned 16-QAM modulation scheme, which leads to the above last assumption.

It is further noted that the variance of x is V{x} = σ \ = 0.5 , as it denotes only the real or imaginary part of a 16-QAM constellation with average power of 1.

With that the PDF of y can be expressed as

Now, the magnitude of y is introduced as a derived variable defined as m = , for which the PDF can be expressed as follows:

P_mCn) =

= P_y(y) + P_y(-y) = 2-_Py(y),m≥0. (8)

The expectation value of m is given by the following expression:

This can be expressed by:

(10)

Now, the real valued amplitude scaling factor a is taken into account, which leads to the expression:

y_k=a(x_k+n_k). (H)

Based on that it can be defined m = y , which results in the expression

E{m} = a ^■ E{m} = (12)

This can be rewritten as follows:

E{m} = a-g(σ), (13)

with

By rearranging equation (13), the following expression for a can be obtained:

While an estimate for E{|j|} can be easily computed at the statistical analyzing unit 22 from the received samples, the noise standard deviation σ is unknown at this point.

The approach proposed here is based on an implicit estimation of the standard deviation σ . First, the variance of y can be expressed as follows:

V{y} = c²-=a²(c_x ² ₊c²) = a²(0.5 + c²). (16)

In order to get rid of the unknown variable a , the variance of y is divided by ι~n² ^{|j|} , i.e. the relative ratio R{p) is defined as:

which is independent of the amplitude scaling a .

By inserting σ in (14) and (17), both the corresponding ratio R (σ ) and the term g (σ ) can be computed. As a result, the unknown noise standard deviation σ is

eliminated and g can be plotted over R (σ ) , i.e. g =

The required argument ^ {j}/ EjJj[I can be easily estimated at the statistical analyzing unit 22 from the received signal. It can be used by the scaling factor estimation unit 26 as an address to the lookup table 25 implementing the function /(.) , thus providing g . Fig. 5 shows a diagram indicating the function g normalized to g(σ = 0) = Vo.4 . The distance between two markers of the curve corresponds to a SNR difference of 1 dB. This normalized function can be interpreted as an amplitude scaling or correction factor, where the argument of the function g is an indirect measure for the noise variance of the unsealed signal. The smallest value of the ratio Fjjj/Ejljl} can take in the absence of noise

(σ = 0) is 1.25. As in a real system only the related estimates E{|j|},F{5>} are available,

values <1.25 can occur and should be mapped to g = Vθ.4 .

The values stored in the LUT 25 are computed once, to be available for all times, since they do not change for a specific modulation scheme. In summary, a receiver apparatus and method of blind amplitude estimation for a received complex signal have been described, wherein the complex signal is analyzed to determine first and second order statistics of real part and imaginary part of the complex signal. Based on a predetermined non-linear function, amplitude scaling information is pre- calculated and used to derive a desired amplitude scaling information for said determined first and second order statistics. Thereby, a low complexity method for blind amplitude estimation for unknown data symbols distorted by an unknown amount of noise can be provided. It is noted that the present invention can be applied to any receiver of any communication system employing an M-ary linear modulation, where the amplitude of the transmitted symbols carries information. Due to the simple and easy processing steps according to the preferred embodiment, the present invention is particularly advantageous for high-speed downlink packet access (HSDPA) as standardized within 3GPP UMTS frequency division duplex (FDD) framework.

The present invention is not restricted to the above calculation of variance and expectation values, and any kind of first order statistics and second order statistics - or even higher order statistics - could be used to obtain easily derivable statistical values which can be mapped via a suitable function to scaling factor estimations or amplitude estimations. The preferred embodiments may thus vary within the scope of the attached claims.

Finally but yet importantly, it is noted that the term "comprises" or

"comprising" when used in the specification including the claims is intended to specify the presence of stated features, means, steps or components, but does not exclude the presence or addition of one or more other features, means, steps, components or group thereof. Further, the word "a" or "an" preceding an element in a claim does not exclude the presence of a plurality of such elements. Moreover, any reference sign does not limit the scope of the claims.

Claims

CLAIMS:

1. A receiver apparatus for receiving a complex signal via a digital transmission channel, said receiver apparatus comprising: analyzing means (22) for analyzing said complex signal to determine first and second order statistics of real part and imaginary part of said complex signal; memory means (25) for storing amplitude scaling information; determination means (26) for addressing said memory means based on said first and second order statistics in order to derive a desired amplitude scaling information; and detecting means (24) for detecting information provided in said complex signal by using said desired amplitude scaling information.

2. An apparatus according to claim 1, wherein said memory means comprise a look-up table (25).

3. An apparatus according to claim 1 or 2, wherein said first and second order statistics relate to an expectation value of the magnitude, and the variance of said complex signal.

4. An apparatus according to any one of the preceding claims, wherein said first and second order statistics are obtained by taking a predetermined number of received symbols of said complex signal into account.

5. An apparatus according to any one of the preceding claims, wherein said determination means (26) is adapted to address said memory means (25) based on a ratio between said first and second order statistics.

6. A method of blind amplitude estimation for a received complex signal, said method comprising the steps of: analyzing said complex signal to determine first and second order statistics of real part and imaginary part of said complex signal; pre-calculating amplitude scaling information based on a predetermined nonlinear function; and using said pre-calculated amplitude scaling information to derive a desired amplitude scaling information for said determined first and second order statistics.

7. A method according to claim 6, wherein said predetermined non-linear function is derived from probability density functions of noise and symbols provided in said complex signal.

8. A method according to claim 6 or 7, wherein said received complex signal is modulated by an M-ary linear modulation.