WO2009118700A1

WO2009118700A1 - System and method for high performance finite-sample-based noise variance estimation for td-scdma

Info

Publication number: WO2009118700A1
Application number: PCT/IB2009/051254
Authority: WO
Inventors: Xun Fan; Liang Wang
Original assignee: Nxp B.V.
Priority date: 2008-03-26
Filing date: 2009-03-25
Publication date: 2009-10-01

Abstract

A network communication system for computing noise variance. The network communication system includes a receiver and a noise variance estimator. The noise variance estimator includes an average power estimator and a noise variance engine. The receiver receives a signal transmitted from a transmitter and processes the received signal. The noise variance estimator estimates a noise variance according to a noise variance algorithm. The average power estimator calculates an average power of a plurality of low-power samples. The noise variance engine derives a final noise variance estimate based on a relationship between an a posteriori mean of noise power and an a priori noise variance. The noise variance engine treats the calculated average power of the plurality of low-power samples as the a posteriori mean of the noise power.

Description

SYSTEM AND METHOD FOR HIGH PERFORMANCE FINITE-SAMPLE- BASED NOISE VARIANCE ESTIMATION FOR TD-SCDMA

TD-SCDMA is an international standard for 3^rd generation mobile communication systems. TD-SCDMA is a central standard for Chinese network communication systems. In TD-SCDMA systems, the noise variance, or Interference Signal Code Power (ISCP) for the single cell detection case, is an important measurement made at the receiver. The accuracy of the noise variance measurement affects at least the performance of active channel window detection, Minimum Mean Square Error Block Equalization (MMSE-BLE) data detection, Dynamic Channel Allocation (DCA), and Radio Resource Management (RRM).

For TD-SCDMA systems, the minimum period for performing noise variance estimation is a timeslot, or timeslot. At each timeslot, a structure -predetermined data burst is transmitted. Fig. 1 depicts a block diagram of one embodiment of a timeslot burst structure 100. TD-SCDMA systems transmit signals according to a specific structure at each timeslot as shown in Fig. 1. A TD-SCDMA timeslot is designed to fit into exactly one burst, and its length is 675us. The timeslot includes four parts. In particular, the timeslot burst structure 100 includes a midamble (144 chips) for channel estimation, two identical data fields (352 chips) at each side of the midamble for bearing user information, followed by a 16 chips guard period (GP) for inter- timeslot interference mitigation. A chip is one bit of a direct-sequence spread spectrum code. The chip rate of a code is the number of bits per second (chips per second) at which the code is transmitted (or received). On the two data parts, the same spreading code channels are allocated. Over each spreading code channel which spans over the two data parts, a sequence of modulation symbols drawn from a constellation are carried after spreading by a specific spreading code. Constellations may include phase-shift keying such as binary phase-shift keying (BPSK), quadrature phase-shift keying (QPSK), high-order PSK such as 8PSK, and quadrature amplitude modulation (QAM) such as 16QAM. A GP represents a first switching point between the downlink and uplink transmission direction. The GP is used to alleviate the effect of multipath delay, associated with radio mobile communications, for frequency flat Rayleigh fading channels, and for two-path fading channels in the presence of additive white Gaussian noise (AWGN). The midamble, or training sequence, is used by the receiver to carry out channel estimation tasks.

Conventional noise variance estimation suffers heavy performance loss due to large estimation bias. Mean Square Error (MSE) performance degradation is due to large estimation bias. The performance of conventional noise variance estimation in the case of L = 64 is substantially reduced, where the normalized estimation bias is

1 given by bias (N₀) D | E{N₀} - N₀ \ , the normalized estimation variance is given

by _Var(jV₀ ) D — — E { \ E{N₀} - N₀ \²} ■> ^and the normalized estimation MSE is given by

N/

MSE(N₀) U -^E{\ E{N₀} - N₀ \²} = var(N₀) + bias(N₀)² .

Embodiments of a system are described. In one embodiment, the system is a network communication system for computing noise variance. Embodiments of the network communication system include a receiver and a noise variance estimator. The noise variance estimator includes an average power estimator and a noise variance engine. The receiver receives a signal transmitted from a transmitter and processes the received signal. The noise variance estimator estimates a noise variance according to a noise variance algorithm. The average power estimator calculates an average power of low-power samples. The noise variance engine derives a final noise variance estimate based on a relationship between an a posteriori mean of noise power and an a priori noise variance. The noise variance engine treats the calculated average power of the plurality of low-power samples as the a posteriori mean of the noise power. Other embodiments of the system are also described.

Embodiments of a method are also described. In one embodiment, the method is a method for computing a noise variance estimate. The noise variance method includes receiving a signal transmitted from a transmitter and processing the received signal. The noise variance method also includes calculating an average power of a plurality of low-power samples of the received signal. The noise variance method also includes designating the calculated average power of the plurality of low-power samples as an a posteriori mean of the noise power. The noise variance method also includes deriving a final noise variance estimate based on a relationship between the a posteriori mean of noise power and an a priori noise variance. Other embodiments of the method are also described.

Other aspects and advantages of embodiments of the present invention will become apparent from the following detailed description, taken in conjunction with the accompanying drawings, illustrated by way of example of the principles of the invention.

Fig. 1 depicts a block diagram of one embodiment of a timeslot burst structure.

Fig. 2 depicts a schematic block diagram of one embodiment of a network communication system.

Fig. 3 depicts a schematic diagram of one embodiment of a receiver for use with a noise variance estimator.

Fig. 4 depicts an overall performance data chart for a Maximum-Sample- Power-Revised Noise Variance Estimation (MSPR_NVE) function/(x). Fig. 5 depicts a specific range of the performance data chart of Fig. 4 for very small values of x ( 0 < x ≤ 0.0007 ).

Fig. 6 depicts another specific range of the performance data chart of Fig. 4 for small values of x ( 0.0007 < x ≤ 0.25 ).

Fig. 7 depicts another specific range of the performance data chart of Fig. 4 for medium values of x ( 0.25 < x ≤ 8).

Fig. 8 depicts another specific range of the performance data chart for the MSPR_NVE function/x) for large values of x ( JC > 8 ).

Fig. 9 depicts a schematic flow chart diagram of one embodiment of a single - timeslot MSPR NVE computation method for use with the noise variance estimator of Fig. 3. Fig. 10 depicts a schematic flow chart diagram of one embodiment of a multi- timeslot MSPR NVE computation method for use with the noise variance estimator of Fig. 3.

Fig. 11 depicts a schematic flow chart diagram of another embodiment of a multi-timeslot MSPR NVE computation method for use with the noise variance estimator of Fig. 3.

Fig. 12 depicts a schematic flow chart diagram of one embodiment of an engineering implementation example for MSPR NVE operation.

Figs. 13A-13C depict performance data charts for a single-timeslot MSPR_NVE computation.

Figs. 14A-14C depict performance data charts for a 7-timeslot MSPR_NVE computation.

Figs. 15 A-15C depict performance data charts for a 14-timeslot MSPR_NVE computation. Throughout the description, similar reference numbers may be used to identify similar elements.

Fig. 2 depicts a schematic block diagram of one embodiment of a network communication system 200. As depicted in Fig. 2, the network communication system 200 includes a base station (BS) 202, at least one antenna 204, a network interface 206, and a user equipment (UE) 208. Although the depicted network communication system 200 is shown and described herein with certain components and functionality, other embodiments of the network communication system 200 may be implemented with fewer or more components or with less or more functionality. For example, some embodiments of the network communication system 200 include at least one base station 202, at least one network interface 206, and at least one UE 208. Additionally, some embodiments of the network communication system 200 include similar components arranged in another manner to provide similar functionality, in one or more aspects.

The base station 202 includes a transmitter 210, a receiver 212, a processor 214, and a memory device 216. The base station 202 connects to the network interface 206 through the antenna 204. In one embodiment, the base station 202 is a radio receiver/transmitter, or transceiver. In some embodiments, the base station 202 is a hub of a local wireless network. In further embodiments, the base station 202 is a gateway between a wired network and a wireless network. In another embodiment, the base station 202 is a wireless communications station installed at a fixed location. In some embodiments, the base station 202 is a wireless cell phone tower and/or wireless data tower.

In one embodiment, the transmitter 210 modulates signals in a multi-mode environment. In some embodiments, the transmitter 210 modulates signals using binary phase-shift keying (BPSK), quadrature phase-shift keying (QPSK), or quadrature amplitude modulation (QAM) and sends a modulated signal to at least one UE 208.

The BS receiver 212 includes a noise variance estimator 218. In one embodiment, the BS receiver 212 receives a signal from at least one UE 208. The BS receiver 212 then demodulates the received signal. In one embodiment, the receiver 212 demodulates the received signal and computes a noise variance associated with the received signal via the noise variance estimator 218.

In one embodiment, the processor 214 is a central processing unit (CPU) with one or more processing cores. In other embodiments, the processor 214 is a network processing unit (NPU) or another type of processing device such as a general purpose processor, an application specific processor, a multi-core processor, or a microprocessor. Alternatively, a separate processor may be coupled to the noise variance estimator 218. In general, the processor 214 executes one or more instructions to provide operational functionality to the base station 202. The instructions may be stored locally in the processor 214 or in the memory device 216. Alternatively, the instructions may be distributed across one or more devices such as the processor 214, the memory device 216, the noise variance estimator 218, or another data storage device.

In some embodiments, the memory device 216 is a random access memory (RAM) or another type of dynamic storage device. In other embodiments, the memory device 216 is a read-only memory (ROM) or another type of static storage device. In other embodiments, the illustrated memory device 216 is representative of both RAM and static storage memory within a network communication system 200. In some embodiment, the memory device 216 is content-addressable memory (CAM). In other embodiments, the memory device 216 is an electronically programmable read-only memory (EPROM) or another type of storage device. Additionally, some embodiments store instructions as firmware such as embedded foundation code, basic input/output system (BIOS) code, and/or other similar code.

In one embodiment, the noise variance estimator 218 implements a high- performance finite-sample-based noise variance estimation to compute a final noise variance. The high-performance finite-sample-based noise variance estimation algorithm, or maximum-sample -power-revised noise variance estimation (MSPR NVE) algorithm, is based on theoretical derivation and engineering approximation. The noise variance estimator 218 first calculates the average power of the available lower-power samples, and then, taking this average power as the a posteriori mean of noise power constrained by the maximum power of available samples, and based on the theoretically-derived relationship between the a posteriori mean of noise power and the a priori noise variance (the a prior noise variance refers to the to-be-estimated variance of additive Gaussian noise at the receiver without any constrained conditions), the noise variance estimator 306 uses MSPR_NVE to derive the final noise variance.

Following the generation of a modulated signal, the base station 202 transmits the signal to at least one UE 208 through the base station antenna 204. In one embodiment, the antenna 204 transmits and/or receives network communications between the base station 202 and at least one UE 208 through the network interface 206. The base station antenna 204 may include an omni-directional antenna, directional antenna, sectoral antenna, panel antenna, and/or other type of antenna. In some embodiments, the base station antenna 204 sends a modulated signal over the air (OTA), through the network interface 206. In other embodiments, the base station 202 sends the modulated signal over physical transmission lines, such as over a coaxial transmission line. Thus, the network interface 206 includes OTA wireless transmissions as well as physical wired transmissions. In a further embodiment, the antenna 204 includes multiple antennas attached to the base station 202, such as the multiple antennas used in multiple- input and multiple-output (MIMO) systems. The UE 208 also includes a transmitter 220, a receiver 222, a processor 224, a memory device 226, and a noise variance estimator 228. In some embodiments, at least some of the components of the UE 208 are substantially similar and operate in a substantially similar manner to the components described above with regard to the BS 202. Additionally, the UE 208 also includes at least one UE antenna 230 and a wired network connection 232. In some embodiments, the UE 208 includes the antenna 230 or the wired network connection 232. The UE 208 may be a desktop computer, a laptop computer, a personal digital assistant (PDA), a cell-phone, a voice-over internet protocol (VoIP) telephone, or other similar client device.

In one embodiment, UE noise variance estimator 228 operates substantially similar to the BS noise variance estimator 218. In one embodiment, the BS transmitter 210 modulates a signal and transmits the modulated signal across the network interface 206 to the UE 208 via the BS antenna 204. The UE receiver 222 receives the modulated signal from the base station 202 and processes the received signal according to the UE noise variance estimator 228. In another embodiment, the UE transmitter 220 modulates a signal and transmits the modulated signal across the network interface 206 to the BS 202 via the UE antenna 230. The BS receiver 212 receives the modulated signal from the UE 208 and processes the received signal according to the operations of the BS noise variance estimator 218. The UE 208 receives and transmits signals through the antenna 230 and/or wired network connection 232. Thus, the transmission and/or reception of a signal may be over a wired connection or over a wireless transmission. Exemplary wired connections include 10/100/1000 BASE-T Ethernet, coaxial cable communications such as Cable Television (CATV) and cable internet, Universal Serial Bus (USB), Institute of Electrical and Electronics Engineers (IEEE) 1394, Recommended Standard 232 (RS-232), etc. Exemplary wireless connections include Wireless- Fidelity (Wi-Fi), WiMAX, 3rd Generation Partnership Project (3GPP), Universal Mobile Telecommunications System (UMTS), High-Speed Packet Access (HSPA), Infrared Data Association (IrDa), Bluetooth, including transport layers implemented over any of Wireless Access Protocol (WAP), Hypertext Transfer Protocol (HTTP), Object Exchange (OBEX), or other similar transports. The depicted components of the base station 202 and/or user equipment 208 may include one or more bus interfaces (not shown) to facilitate communications related to performing MSPR_NVE with the BS noise variance estimator 218 and/or UE noise variance estimator 228, as well as storing, sending, and receiving data packets associated with the MSPR NVE computation.

Fig. 3 depicts a schematic diagram of one embodiment of a receiver 300 for use with a noise variance estimator 306. In some embodiments, the receiver 300 is substantially similar to the BS receiver 212 and/or the UE receiver 222 with reference to Fig. 2. The receiver 300 includes a received signal sequencer 302, a channel estimation module 304, a noise variance estimator 306, a post processing module 308, a lookup table 310, and one or more bus interfaces 320 to facilitate communications related to a noise variance computation executed on the noise variance estimator 306, including processing noise variance computation commands, as well as storing, sending, and receiving data packets associated with the noise variance estimator 306. Although the depicted receiver 300 is shown and described herein with certain components and functionality, other embodiments of the receiver 300 may be implemented with fewer or more components or with less or more functionality. For example, some embodiments of the receiver 300 includes at least one received signal sequencer 302, at least one channel estimation module 304, at least one noise variance estimator 306, and at least one post processing module 308. Additionally, some embodiments of the receiver 300 include similar components arranged in another manner to provide similar functionality, in one or more aspects.

In one embodiment, the received signal is first received by the received signal sequencer 302. The received signal sequencer 302 separates a received signal into a data related received sequence and a midamble related received sequence. The midamble related received sequence is sent to the channel estimation module 304 in order to obtain the estimate of channel state information. The channel estimation module 304 generates and sends a raw channel response estimation to the noise variance estimator 306.

In one embodiment, the noise variance estimator 306 includes an average power estimator 312, a noise variance engine 314, and a memory device 316 that stores a noise variance algorithm 318. The noise variance estimator 306 performs a high-performance finite-sample-based noise variance estimation operation to compute noise variance. The high-performance finite-sample-based noise variance estimation, otherwise known as maximum-sample -power-revised noise variance estimation (MSPR NVE), is based on theoretical derivation and engineering approximation. In one embodiment, the average power estimator 312 calculates the average power of the available low-power samples. In some embodiments, the noise variance engine 314 takes the average power calculated by the average power estimator 312 as the a posteriori mean of noise power constrained by the maximum power of available samples. Based on the theoretically-derived relationship between the a posteriori mean of noise power and the a priori noise variance, the noise variance engine 314 derives the final noise variance estimate. In some embodiments, the noise variance algorithm is an MSPR_NVE algorithm.

MSPR_NVE substantially reduces an estimation bias in the finite sample case, and hence, improves MSE performance. Moreover, MSPR_NVE performance may be further improved through multi-timeslot based joint noise variance estimation. In one embodiment, the noise variance estimator 306 refers to a lookup table 310 to perform an MSPR NVE computation according to a lookup-table based function approximation.

At the receiver 300, the received signal sequencer 302 separates a data sequence and a midamble sequence e_m = [e_{m 17} , e_{m 18} , ..., e_{m 144} ]^τ from the received signal. In some embodiments, the midamble sequence e_m is composed of the last P = 128 samples of the 144 chip midamble. Assuming the channel length is less than 16, e_m is modeled as: e_m = Gh + n (1) where G is a circulant matrix with the first column (the vector of the basic midamble in the concerned cell), h = [h_l,h₂,...,h_Pf is the joint channel response to be estimated, and n is the AWGN vector at the receiver 300 with the entries modeled as independent and identically distributed (i.i.d.) Gaussian random variables with zero mean and unknown variance No- Thus, based on the e_m model and the a priori basic midamble knowledge, the channel estimation module 304 derives the Maximum Likelihood (ML) estimate of h , called the raw channel response estimate h = [h_i,h₂,...,h_pf , as:

Because of the circulant structure of G , the ML estimation in (1) can be implemented via a low-complexity FFT computation as: h = IFFT (FFT (e Jl FFT '(m)) (2) where FFT and IFFT denote the Fast Fourier Transform and Inverse Fast Fourier Transform respectively, and m denotes the P-dimensional basic midamble vector.

In the raw channel response estimate vector h , some samples include both a non- zero-power channel impulse response component and a noise component. Some only include the noise component with variance N₀ ' = — N₀ ,

where (FFr(In)I₁ denotes the operation of drawing the z^th entry of sequence FFT(m). Based on (4), the noise variance estimator 306 derives the estimation of N₀ , N₀ , at a timeslot through the following operations. In a first operation, the noise variance estimator 306 selects L low-power samples from the P samples in the output raw channel response estimation h , (S₁, S₂, ..., S_L). Sorting the samples of h in an ascending order according to their power level, the noise variance estimator 306 generates a new sequence h^(s) = sort(h) = \h , h , ..., h 1 , where h ²≤ h ²≤ ... ≤\ h ² , and the L low-power samples (S^s₂, ...,s_L) are the first L samples in h^(s) . The noise variance estimator 306 draws from h the L samples including a noise-only component for the subsequent variance estimation operation. In one embodiment, Z is a predetermined parameter. In some embodiments, when the number of expected non-zero-power channel taps drops below a certain expected channel tap threshold, the noise variance estimator 306 selects a larger L value (i.e., approaching P = 128 ; for example, going from L = 64 to L = 100) to increase estimation accuracy.

In a second operation, the noise variance estimator 306 derives the estimate of N₀ , N₀ , through power averaging of the L low-power samples. That is,

N₀ = - N₀ ' = - Y \ s |² (4)

⁰ D ° DL tr ' The noise in multiple neighboring receiving timeslots may be considered to possess similar statistical characteristics. A multi-timeslot joint noise variance operation averages the output of all single -timeslot noise variance estimations, and may be implemented to substantially increase performance.

The received signal, the raw channel estimation response estimate h , data sequence, and noise variance estimate N₀ are implemented as the input of the subsequent post processing module 308 of the receiver 300 to generate resultant signal information data.

The noise samples in the output sequence of channel estimation (i.e., the noise- only component entries of h in (2)) are nearly i.i.d. Gaussian random variables with zero-mean and variance ^v 0 ' = — p N 0 ■ Thus, when the number of available samples for noise variance estimation, or equivalently, the number of selected low-power samples in the raw channel response estimate vector at each timeslot, is large (e.g., L = 100 ) the power-average based noise variance estimation method performance is substantially improved. However, in view of the large number of expected non-zero- power-channel-tap samples in the raw channel response estimate vector, L is selected as a relatively small value (e.g., L = 64 ) to ensure a proper input to the subsequently selected sample -based noise variance estimation.

The theoretical foundation of the proposed high performance finite-sample- based noise variance estimation method, MSPR_NVE, is the insight on a relationship between .E(N² 1 N² < a} and N₀ , i.e.,

£{N² | N² ≤ α} = + N₀ (5) exp( — ) - 1

N or equivalently,

^^■ | Ν' ≤ -} ₌ l__{+ i > χ = JL (6)} a exp(x) - l x N₀ where E[N² | N² < a} is the a posteriori expectation of the power of a zero-mean N₀ - variance Gaussian random variable (the power denoted as the random variable Ν² ), and is conditioned on the event that only samples with power below a certain value a in the overall Gaussian random variable sample space are observed. The event is denoted by Ν² ≤ a . N₀ is the a priori variance of the Gaussian random variable. In some embodiments, Eq. 5 and/or Eq. 6 define the MSPR ΝVE algorithm. In some embodiment, the noise variance algorithm 318 is defined by Eq. 5. In some embodiments, the noise variance algorithm 318 is defined by Eq. 6. Other embodiments may use other forms of the MSPR_ΝVE algorithm.

Based on the theoretical insight given above, in one embodiment, the noise variance estimator 306 selects a certain number of low-power samples from the output of the channel estimation module 304 at the corresponding timeslot. The noise variance estimator 306 calculates the average power for the available low-power samples of the corresponding timeslot. In some embodiments, the noise variance estimator 306 performs an MSPR NVE operation to obtain the final noise variance estimate. In other words, the noise variance estimator 306 takes the average power as the a posteriori mean of noise power constrained by the maximum power of available samples, and based on the theoretically-derived relationship between the a posteriori mean of noise power and the a priori noise variance, the noise variance estimator 306 computes the final noise variance estimate. For multi-timeslot joint noise variance estimation, in some embodiments, the noise variance estimator 306 performs one of at least two sub-types of MSPR NVE according to the granularity of performing MSPR_NVE operation as follows:

MSPR-based joint estimation type-I: For MSPR NVE joint estimation type-I, in one embodiment, the noise variance estimator 306 selects a group of timeslots and performs independent MSPR NVE operations for each of the timeslots. For each timeslot, the noise variance estimator 306 selects a certain number of low-power samples from the output of corresponding channel estimation from the channel estimation module 304. For each timeslot, average power estimator 312 calculates the average power of the available low-power samples. The noise variance engine 314 then performs an MSPR_NVE operation according to the noise variance algorithm 318 on the selected samples to derive a corresponding noise variance estimate value of each timeslot. The noise variance engine 314 then averages the corresponding noise variance estimate of each timeslot to obtain a final multi-timeslot joint noise variance. MSPR-based joint estimation type-II: For MSPR NVE joint estimation type-

II, in one embodiment, the noise variance estimator 306 jointly selects a certain number of low-power samples from the output of the channel estimation module 304 corresponding to a group of timeslots. The average power estimator 312 calculates the average power of the available low-power samples. The noise variance engine 314 then performs an MSPR_NVE operation according to the noise variance algorithm 318 on the jointly-selected samples to obtain the final multi-timeslot joint noise variance.

Letting N be a zero-mean TV₀ -variance complex Gaussian random variable, where the overall sample space of which is denoted as D , Y = N² be a random variable with the corresponding sample space D ² composed of the power of the samples of N (the term "power" for a sample meaning the square of modulus of the sample), Y = N² ≤ a be an event which denotes a sub-sample space of D ² with sample value not greater than a certain value a (equivalently, a sub-sample space of D with the sample power being not greater than a ) is observed. The a posteriori estimate of Y given the event Y < a is known to have occurred, i.e., E{Y | Y < a} , can be obtained via:

£{Y | Y ≤ α} = £{N² | N² ≤ α} = + N₀ (?) exp( — ) - 1

N

Since N obeys Gaussian distribution, Y = N² obeys the X² -distribution as:

Given the event Y < a is known to have occurred, the a posteriori probability density of Y can be obtained by Bayes' theorem as:

fi, y ≤ a where l{y ≤ a} Ω <

{ 0, else Thus, ^E(Y I Y < a} , can be calculated as: .E(Y I Y < a} = E{W I N² < a}

Eq. 8 demonstrates the relationship between the a posteriori expectation of power of a zero-mean N₀ -variance Gaussian random variable conditioned on the event that only the samples with power levels below a certain value a in the overall Gaussian random variable sample space is observed, E[N² | N² < a} , and the a priori variance of the concerned Gaussian random variable, N₀ . Such as relationship is a component of embodiments the MSPR ΝVE operation for obtaining a final noise variance value. An alternative form of Eq. 8 may be given as:

E[Y I Y < a} £{Ν² 1 N² < a} 1 1 a

^■ = /(*) = exp T(xV) -TI ⁺-x ' * = 7 NT₀ ^{( 1 1)}

In some embodiments, Eq. 7 and/or Eq. 11 define the MSPR NVE algorithm. In some embodiments, the noise variance algorithm 318 is defined by Eq. 7. In some embodiments, the noise variance algorithm 318 is defined by Eq. 11.

Fig. 4 depicts an overall performance data chart 400 for a Maximum-Sample-Power- Revised Noise Variance Estimation (MSPR_NVE) functionyζλ;). In particular, Fig. 4 shows performance simulations of the above described MSPR_NVE computations, which include Eq. 11. Fig. 4 illustrates an overall perspective of the function f(x) in Eq. 11.

Fig. 5 depicts a specific range of the performance data chart 500 of Fig. 4 for very small values of x ( 0 < x ≤ 0.0007 ). In particular, Fig. 5 shows performance simulations ofy(x) for very small values of x (0 < x ≤ 0.0007) corresponding to 0.4999417 < f(x) < 0.5 . Fig. 5 depicts the high frequency oscillation inf(x) for very small values of x. The oscillations depicted in Fig. 5 demonstrate that there are multiple values ofx corresponding to the same/(x) value. Thus, finding an x value from a known f(x) value in the very small x region is inconclusive. Fig. 6 depicts another specific range of the performance data chart 600 of Fig.

4 for small values ofx (0.0007 < x ≤ 0.25). In particular, Fig. 6 shows performance simulations of/(x) for small values ofx (0.0007 < x ≤ 0.25) corresponding to 0.4792 < f(x) < 0.4999417 . Fig. 6 does not depict any obvious oscillation corresponding to f(x) . In the small x region, f(x) may be approximated as a decreasing function. However, since the slope of/(x) in the small x region is nearly zero, finding a precise value ofx from a known /(x) value is difficult through engineering approximation methods (e.g., a lookup-table based function approximation associated with the lookup table 310).

Fig. 7 depicts another specific range of the performance data chart 700 of Fig. 4 for medium values of x ( 0.25 < x ≤ 8 ). In particular, Fig. 7 shows performance simulations of/(x) for medium values ofx (0.25 < x ≤ 8) corresponding to 0.1247 < f(x) < 0.4792. In the medium x region/(x) may also be approximated as a decreasing function. Finding an x value from a known /(x) value associated with the medium x region is derived with engineering approximation methods. In one embodiment, the noise variance estimator 306 refers to the lookup table 310 to perform lookup-table based function approximations associated with the MSPR NVE computation.

Fig. 8 depicts another specific range of the performance data chart 800 for the MSPR_NVE function/(x) for large values of x (x > 8 ). In particular, Fig. 8 shows performance simulations of/(x) for large values ofx (x > 8 ) corresponding to f(x) < 0.1247 . Finding anx value from a known f(x) value may be approximated by an inverse proportion function g(x) = 1/x in the large x region. The relationship shown by Eq. 12 and the characteristics of /(x) therein make it feasible to find the a priori noise variance N₀ from the a posteriori expectation is {Y I Y ≤ a} by implementing an engineering approximation operation and/or direct- calculation operation for the majority of E{Y | Y < a} values (e.g. f(_x) = ^Ei^Y \ ^{Y < a}} < _{0A792 s or} correspondingly, x > 0.25 . a

Fig. 9 depicts a schematic flow chart diagram of one embodiment of a single - timeslot MSPR NVE computation method 900 for use with the noise variance estimator 306 of Fig. 3. Although the single-timeslot MSPR NVE computation method 900 is described in conjunction with the noise variance estimator 306 of Fig. 3, some embodiments of the method 900 may be implemented with other types of noise variance estimators. At block 902, the average power estimator 312 selects a certain number of low-power samples from the output of the channel estimation module 304 corresponding to a single timeslot. At block 904, the average power estimator 312 calculates the average power of the available low-power samples. At block 906, the noise variance engine 314 performs an MSPR NVE operation to obtain a final noise variance estimate corresponding to the single timeslot.

For the multi-timeslot joint noise variance estimation, the proposed method can be generalized into two sub-types of methods according to the granularity of performing MSPR NVE operation as follows with reference to Figs. 10 and 11. Fig. 10 depicts a schematic flow chart diagram of one embodiment of a multi-timeslot MSPR NVE computation method 1000 for use with the noise variance estimator 306 of Fig. 3. Although the multi-timeslot MSPR NVE computation method 1000 is described in conjunction with the noise variance estimator 306 of Fig. 3, some embodiments of the method 1000 may be implemented with other types of noise variance estimators. At block 1002, the average power estimator 312 selects a single timeslot from a group of timeslots. At block 1004, the average power estimator 312 selects a certain number of low-power samples from the output of the channel estimation module corresponding to the selected timeslot. At block 1006, the average power estimator 312 calculates the average power of the available low-power samples. At block 1008, the noise variance engine 314 performs an MSPR NVE operation to obtain the resultant noise variance value of the selected timeslot. At block 1010, the noise variance estimator 306 determines whether each of the timeslots from the group of timeslots is processed. If the noise variance estimator 306 determines that at least one of the timeslots of the group of timeslots remains unprocessed, then the method 1000 returns to block 1002 to select the next timeslot for processing. Otherwise, if the noise variance estimator 306 determines that all of the timeslots of the group of timeslots are processed, the method 1000 proceeds to block 1012. At block 1012, the noise variance engine 314 averages the resultant noise variance estimates corresponding to each of the processed timeslots from the group of timeslots to obtain the overall final noise variance estimate corresponding to the group of timeslots.

In other words, the method 1000 performs independent MSPR NVE for each of the selected timeslots in a group of timeslots to obtain corresponding noise variance estimation values. For each timeslot, the average power estimator 312 selects a certain number of low-power samples from the output of corresponding channel estimation (e.g., L = 64 low-power samples from a total P = 128 output samples of channel estimation from the channel estimation module 304), calculates the average power of the available lower-power samples, and then performs the MSPR_NVE operation on the selected samples to obtain the corresponding noise variance estimate value. Lastly, the noise variance estimator 306 averages on the corresponding noise variance estimate values to obtain a final multi-timeslot joint noise variance estimate. Fig. 1 1 depicts a schematic flow chart diagram of one embodiment of another multi-timeslot MSPR NVE computation method 1 100 for use with the noise variance estimator 306 of Fig. 3. Although the other multi-timeslot MSPR NVE computation method 1 100 is described in conjunction with the noise variance estimator 306 of Fig. 3, some embodiments of the method 1 100 may be implemented with other types of noise variance estimators. At block 1 102, the average power estimator 312 selects a certain number of low-power samples from the output of the channel estimation module 304 corresponding to a group of timeslots. At block 1104, the average power estimator 312 calculates the average power of the available low-power samples. At block 1106, the noise variance engine 314 performs an MSPR_NVE operation to obtain the overall final noise variance estimate corresponding to the group of timeslots

In other words, the average power estimator 312 jointly selects a certain number of low-power samples from the output of channel estimation of all the selected timeslots in a group of timeslots (e.g. LT = 64-T lower-power samples from total PT = \28T output samples of channel estimation, where T is the number of selected timeslots). The average power estimator 312 then calculates the average power of the available lower-power samples, and performs the MSPR_NVE operation on the jointly-selected samples to obtain the final multi-timeslot joint noise variance estimate. Fig. 12 depicts a schematic flow chart diagram of one embodiment of an engineering implementation method 1200 for MSPR NVE operation. Although the engineering implementation method 1200 is described in conjunction with the noise variance estimator 306 of Fig. 3, some embodiments of the method 1200 may be implemented with other types of noise variance estimators. At block 1202, the noise variance estimator 306 determines whether the estimate of the posterior expectation, E{Y | Y < a} , in relation to a, a predetermined threshold value, falls within the following range: 0.1247 < E{Y | Y < a} I a < 0.4792 . The designated range corresponds to the function region of f(x) with medium x value (particularly, 0.25 < x ≤ 8). Therefore, when 0.1247 < E{Y I Y < a) I a < 0.4792 , then at block 1204, the average power estimator 312 calculates x = f^~l(E{Y | Y < a] I a) in conjunction with the lookup table 310 for the approximation of f(x) . At block 1206, the noise variance estimate, TV₀ ' , is

obtained by N₀ ' = - . Otherwise, at block 1208, the noise variance estimate, N₀ ' , is x obtained by N₀ ' = .E(Y | Y < a) . The result of N₀ ' = E{Y | Y < a) implies two different physical meanings, respectively for the cases of E{Y | Y < a} I a ≥ 0.4792 and E{Y \ Y ≤ a}/a < 0.1247 .

For the case of E{Y \ Y ≤ a} I a ≥ 0.4792 , which corresponds to the function region of f(x) with very small x value (particularly, 0 < x ≤ 0.0007 ) and small x value (particularly, 0.0007 < x ≤ 0.25 ), precisely calculating x = f^~l (E {Y I Y < a} I a) value is difficult in an engineering implementation. In this case, the result N₀ ' = E{Y | Y < a) simply takes the input E{Y | Y < a} , i.e. the power average of the available samples, as the output noise variance estimate without revising.

For the case of E{Y \ Y ≤ a} I a < 0.1247 , which corresponds to the function region of f(x) with very large x value (particularly, x > 8 ) where f(x) is well approximated by g(x) = 1 / x , x can be calculated by 1/(£{Y | Y < a} I a) , thus N₀ '

is obtained by N₀ ' = - = E{Y \ Y ≤ a} . x Both of the above cases maintain the same noise variance estimation results as that in the traditional method. In one embodiment, after obtaining the estimate of the noise variance after channel estimation, N₀ ' , the resultant noise variance estimate is p obtained directly by N₀ = — N₀ '.

Figs. 13A-13C depict performance data charts for single -timeslot MSPR ΝVE computations. In particular, Figs. 13 A-13C show performance simulations of the above described MSPR ΝVE methods, which include the single- timeslot MSPR ΝVE computation method 900. Figs. 13A-13C depicts the performance of the single-timeslot MSPR-based noise variance estimation compared to conventional single-timeslot noise variance estimation with reference to the normalized estimation bias 1300, the normalized estimation variance 1302, and the normalized estimation Mean Square Error (MSE) 1304. The normalized estimation bias is given by: bias(N₀) U ^- \ E{N₀} -N₀ \ ,

the normalized estimation variance is given by: var(#₀) D -^EiI E[N₀J - NJ²) , and the normalized estimation MSE is given by:

MSE(N₀) D -±_jE{\ E{N₀} -N₀ |²} = var(N₀) + bias(N_of .

Figs. 14A-14C depict performance data charts for 7-timeslot MSPR_NVE computations. In particular, Figs. 14A-14C shows performance simulations of the above described MSPR NVE methods, which include the multi-timeslot

MSPRJSfVE computation methods 1000 and 1100. Figs. 14A-14C depict the performance of the multi-timeslot MSPR-based noise variance estimation for 7- timeslots compared to conventional multi-timeslot noise variance estimation with reference to the normalized estimation bias 1400, the normalized estimation variance 1402, and the normalized estimation Mean Square Error (MSE) 1404.

Figs. 15A-15C depict performance data charts for 14-timeslot MSPR_NVE computations. In particular, Figs. 15A-15C shows performance simulations of the above described MSPR NVE methods, which include the multi-timeslot MSPR NVE computation methods 1000 and 1100. Figs. 15A-15C depict the performance of the multi-timeslot MSPR-based noise variance estimation for 14- timeslots compared to conventional multi-timeslot noise variance estimation with reference to the normalized estimation bias 1500, the normalized estimation variance 1502, and the normalized estimation Mean Square Error (MSE) 1504.

In Figs. 13-15, simulation results are shown to compare the performance of the proposed MSPR_NVE methods with that of the conventional noise variance estimation for single and multiple timeslot cases. As can be seen from the Figs. 13- 15, compared with the conventional noise variance estimation, MSPR-based noise variance estimation effectively reduces an estimation bias of noise variance in the finite sample case, and hence obtains better MSE performance. Moreover, the performance of the single -timeslot MSPR-based noise variance estimation method 900 is further improved through performing multi-timeslot MSPR-based joint noise variance estimation methods 1000 and 1100. It should also be noted that at least some of the operations for the methods may be implemented using software instructions stored on a computer useable storage medium for execution by a computer. As an example, an embodiment of a computer program product includes a computer useable storage medium to store a computer readable program that, when executed on a computer, causes the computer to perform operations, including an operation to receive a signal transmitted from a transmitter and to process the received signal, an operation to select a predetermined number of low-power samples from available samples in an output raw channel response estimation generated from a channel estimation module, and an operation to perform a maximum sample power revised algorithm based on a calculated average power of low-power samples to derive a final noise variance estimate.

Embodiments of the invention can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment containing both hardware and software elements. In one embodiment, the invention is implemented in software, which includes but is not limited to firmware, resident software, microcode, etc.

Furthermore, embodiments of the invention can take the form of a computer program product accessible from a computer-usable or computer-readable storage medium providing program code for use by or in connection with a computer or any instruction execution system. For the purposes of this description, a computer-usable or computer readable storage medium can be any apparatus that can store the program for use by or in connection with the instruction execution system, apparatus, or device.

The computer-useable or computer-readable storage medium can be an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system (or apparatus or device), or a propagation medium. Examples of a computer-readable storage medium include a semiconductor or solid state memory, magnetic tape, a removable computer diskette, a random access memory (RAM), a read-only memory (ROM), a rigid magnetic disk, and an optical disk. Current examples of optical disks include a compact disk with read only memory (CD-ROM), a compact disk with read/write (CD-R/W), and a digital video disk (DVD).

An embodiment of a data processing system suitable for storing and/or executing program code includes at least one processor coupled directly or indirectly to memory elements through a system bus such as a data, address, and/or control bus. The memory elements can include local memory employed during actual execution of the program code, bulk storage, and cache memories which provide temporary storage of at least some program code in order to reduce the number of times code must be retrieved from bulk storage during execution.

Input/output or I/O devices (including but not limited to keyboards, displays, pointing devices, etc.) can be coupled to the system either directly or through intervening I/O controllers. Additionally, network adapters also may be coupled to the system to enable the data processing system to become coupled to other data processing systems or remote printers or storage devices through intervening private or public networks. Modems, cable modems, and Ethernet cards are just a few of the currently available types of network adapters. Although the operations of the method(s) herein are shown and described in a particular order, the order of the operations of each method may be altered so that certain operations may be performed in an inverse order or so that certain operations may be performed, at least in part, concurrently with other operations. In another embodiment, instructions or sub-operations of distinct operations may be implemented in an intermittent and/or alternating manner.

Although specific embodiments of the invention have been described and illustrated, the invention is not to be limited to the specific forms or arrangements of parts so described and illustrated. The scope of the invention is to be defined by the claims appended hereto and their equivalents.

Claims

What is claimed is:

1. A network communication system comprising: a receiver to receive a signal transmitted from a transmitter and to process the received signal; a noise variance estimator coupled to the receiver, the noise variance estimator to estimate a noise variance according to a noise variance algorithm, wherein the noise variance estimator comprises: an average power estimator to calculate an average power of a plurality of low-power samples; a noise variance engine coupled to the average power estimator, the noise variance engine to derive a final noise variance estimate based on a relationship between an a posteriori mean of noise power and an a priori noise variance, wherein the a posterior mean of the noise power comprises the calculated average power of the plurality of low-power samples.

2. The network communication system of claim 0, further comprising a channel estimation module coupled to the noise variance estimator, the channel estimation module to output a raw channel response estimation, wherein the noise variance estimator is further configured to select a predetermined number, L, of low-power samples from P available samples in the raw channel response estimation, wherein the L low-power samples correspond to a single time slot.

3. The network communication system of claim 0, wherein the noise variance estimator is further configured to perform the noise variance algorithm based on the L low-power samples selected from the single time slot.

4. The network communication system of claim 0, further comprising a channel estimation module coupled to the noise variance estimator, the channel estimation module to output a raw channel response estimation, wherein the noise variance estimator is further configured to select a predetermined number, L, of low-power samples from P available samples in the raw channel response estimation, wherein the L low-power samples correspond to a plurality of time slots.

5. The network communication system of claim 0, wherein the noise variance estimator is further configured to perform the noise variance algorithm based on the L low-power samples selected from the plurality of time slots.

6. The network communication system of claim 0, wherein the noise variance estimator is further configured to increase the number, L, of low-power samples from the P available samples in the raw channel response estimation in response to a determination that a number of expected non-zero-power channel taps drops below a predetermined expected channel tap threshold.

7. The system of claim 0, further comprising a memory device coupled to the receiver, the memory device to store the noise variance algorithm, wherein the noise variance algorithm is defined by:

where E{ N²| N² ≤ a} denotes an a posteriori expectation of a power of a zero-mean No-variance Gaussian random variable, Ν denotes the power as a random variable, and No denotes an a priori noise variance of the affected Gaussian random variable.

8. The network communication system of claim 0, wherein the memory device is further configured to store a factor of the noise variance engine, wherein the factor of the noise variance engine comprises . exp(x) - 1

9. The network communication system of claim 0, wherein the memory device is further configured to store a result of the noise variance algorithm.

10. The network communication system of claim 0, further comprising a lookup table coupled to the noise variance estimator, the lookup table to perform noise variance algorithm computation according to a lookup-table based function approximation.

11. A noise variance method comprising: receiving a signal transmitted from a transmitter; calculating an average power of a plurality of low-power samples of the received signal; designating the calculated average power of the plurality of low-power samples as an a posteriori mean of the noise power; and deriving a final noise variance estimate based on a relationship between the a posteriori mean of noise power and an a priori noise variance.

12. The noise variance method of claim 0, further comprising selecting a predetermined number, L, of low-power samples fromi³ available samples in a raw channel response estimation, wherein the L low-power samples correspond to a single time slot.

13. The noise variance method of claim 0, further comprising computing the final noise variance estimate according to a noise variance algorithm based on the L low- power samples selected from the single time slot.

14. The noise variance method of claim 0, further comprising selecting a predetermined number, L, of low-power samples from P available samples a raw channel response estimation wherein the L low-power samples correspond to a plurality of time slots.

15. The noise variance method of claim 0, further comprising computing the final noise variance estimate according to a noise variance algorithm based on the L low- power samples selected from the plurality of time slots.

16. The noise variance method of claim 0, further comprising increasing the number, L, of low-power samples from the P available samples in output raw channel response estimation in response to a determination that a number of expected nonzero-power channel taps drops below a predetermined expected channel tap threshold.

17. The noise variance method of claim 0, further comprising defining the noise variance algorithm by:

E{N² 1 N² < a}

- + — , x = — a exp(X) -l x N₀

where E{ N | N < a} denotes an a posteriori expectation of a power of a zero-mean No-variance Gaussian random variable, Ν² denotes the power as a random variable, and No denotes an a priori noise variance of the affected Gaussian random variable.

18. The noise variance method of claim 0, further comprising storing the noise variance algorithm, a factor of the noise variance engine, and results of the noise variance algorithm.

19. A computer program product comprising a computer useable storage medium to store a computer readable program that, when executed on a computer, causes the computer to perform operations comprising: receive a signal transmitted from a transmitter; calculate an average power of a plurality of low-power samples of the received signal; designate the calculated average power of the plurality of low-power samples as an a posteriori mean of the noise power; and derive a final noise variance estimate based on a relationship between the a posteriori mean of noise power and an a priori noise variance.

20. The computer program product of claim 0, wherein the computer readable program, when executed on the computer, causes the computer to perform operations to define the noise variance algorithm by:

£{N² 1 N² < a} 1

- + — , x = - a exp(X) -l x N₀

where E{ N²| N² < a) denotes an a posteriori expectation of a power of a zero-mean No-variance Gaussian random variable, Ν denotes the power as a random variable, and No denotes an a priori noise variance of the affected Gaussian random variable.