### Table 1: Common polynomial models The conversion from state space to polynomial model is achieved by using the companion parametrisation (or observer canonical parametrisation) of the state space model as described in [16, 21]. Consider the state space system in forward innovations form with matrices de ned in companion parametrisation as below. The dependence is dropped for convenience.

"... In PAGE 51: ...18 50.44 Table1 0: Results when analysing clean and noisy voiced speech. Prediction and Kalman lter innovations are analysed for sum-squared error, whiteness and zero- mean, with median averages over all frames considered.... In PAGE 58: ... Refer to [31, 32, 40] for derivations and [29] for an overview. Criterion Optimisation Citation 1a = arg minjjYp ? ?v;i( )Xv;p( )jj2 F [40] 1b = arg max trf ?v;i( ) ^ RY Y g; Xv;p = ?y v;iYv;p [40] 2 = arg min trf?v;i( )H ^ Un ^ UH n ?v;i( ) ^ RXX( )g [31] Table1 1: DOA Maximum Likelihood Estimation Criteria Matrices are de ned as ^ RXX( ) = 1 N Xv;pXH v;p is the sample correlation matrix of the state sequence. ^ RY Y = 1 N YpYH p is the sample correlation matrix of the output.... In PAGE 60: ...05 69.64 Table1 2: Results when analysing clean voiced speech. Prediction and Kalman lter innovations are analysed for sum-squared error, whiteness and zero-mean, with median averages over all frames presented.... In PAGE 61: ...33 11.24 Table1 3: Results when analysing clean voiced speech. log10 mean squared spectral di erences between noise model spectrograms.... In PAGE 62: ...79 63.12 Table1 4: Results when analysing noisy voiced speech. Prediction and Kalman lter innovations are analysed for sum-squared error, whiteness and zero-mean, with median averages over all frames presented.... In PAGE 63: ...08 12.07 Table1 5: Results when analysing noisy voiced speech. log10 mean squared spectral di erences between noise model spectrograms.... In PAGE 63: ...18 56.90 Table1 6: Results when analysing clean non-voiced speech. Prediction and Kalman lter innovations are analysed for sum-squared error, whiteness and zero-mean, with median averages over all frames presented.... In PAGE 64: ...41 10.41 Table1 7: Results when analysing clean non-voiced speech. log10 mean squared spectral di erences between noise model spectrograms.... In PAGE 64: ...33 49.97 Table1 8: Results when analysing noisy non-voiced speech. Prediction and Kalman lter innovations are analysed for sum-squared error, whiteness and zero-mean, with median averages over all frames presented.... In PAGE 64: ...94 10.94 Table1 9: Results when analysing noisy non-voiced speech. log10 mean squared spectral di erences between noise model spectrograms.... ..."

### Table 3. Architectural characteristics of the system we model.

2002

"... In PAGE 8: ...cribed in Section 3.2. Contention is accurately modeled in the en- tire system, including the busses, the network and the main mem- ory. Table3 lists the main characteristics of the architecture. Applications.... ..."

Cited by 45

### Table 3. Architectural characteristics of the system we model.

in ReVive: Cost-Effective Architectural Support for Rollback Recovery in Shared-Memory Multiprocessors

2002

"... In PAGE 8: ...cribed in Section 3.2. Contention is accurately modeled in the en- tire system, including the busses, the network and the main mem- ory. Table3 lists the main characteristics of the architecture. Applications.... ..."

### Table 3. Architectural characteristics of the system we model.

"... In PAGE 8: ...cribed in Section 3.2. Contention is accurately modeled in the en- tire system, including the busses, the network and the main mem- ory. Table3 lists the main characteristics of the architecture. Applications.... ..."

### Table 1. System Architecture Parameters

1998

"... In PAGE 2: ...2 Model Parameters Model parameters can be classified as either describing the system or describing the application. Table1 defines the system parameters, while Table 2 summarizes the application parameters. From the parameters in Table 2, we can compute the probabilities of the protocol transactions in Table 3.... ..."

Cited by 1

### Table 1. System Architecture Parameters

"... In PAGE 2: ...2 Model Parameters Model parameters can be classified as either describing the system or describing the application. Table1 defines the system parameters, while Table 2 summarizes the application parameters. From the parameters in Table 2, we can compute the probabilities of the protocol transactions in Table 3.... ..."

### Table 1. System Architecture Parameters

"... In PAGE 3: ... Consequently, they do not reserve MSHRs or other cache resources. 3 Model Parameters Table1 defines the system architecture parameters, in- cluding the values that are used in the baseline architecture in Section 5. Occupancies are in units of CPU cycles.... ..."

### Table 1. System Architecture Parameters

"... In PAGE 3: ... Consequently, they do not reserve MSHRs or other cache resources. 3 Model Parameters Table1 defines the system architecture parameters, in- cluding the values that are used in the baseline architecture in Section 5. Occupancies are in units of CPU cycles.... ..."