WCDMA 系統中以EM 演算法實現的SIR 估測法

Abstract

WCDMA 系統中，功率控制(PC)依據SIR 調整傳送功率來對抗快速的通道衰減。所以需要一個快速可靠的方法來估測SIR。大多數的SIR 測量方法都使用解調(despread)後的pilot 位元來估測訊號、干擾和雜訊的功率。本論文把SIR 測量視作高斯混合的問題，並以最大期望值(EM)和統計特性估算SIR。此方法不只可以使用pilot 位元也可使用一般的資料，故可在同依時間內得到較多的資訊來估算SIR。並可得到較快的收斂速度。此外，我們希望減少samples 的數目以增加SIR 更新的速率，故使用了一個Kalman filter 來減輕sample 數目減少為估測器帶來的誤差。 實驗結果顯示，此方法提供了較快的收斂速度且沒有喪失太多的精準度。

In WCDMA systems, power control algorithms are used to cope with the fast channel fading. The signal to interference (SIR) ratio is a common measure for the channel condition. Thus, a fast and accurate SIR estimation method is required. Most present SIR estimation methods use pilot bits to estimate the power of interference and noise. In this thesis, we formulate the SIR measurement problem as a Gaussian mixture identification problem. Applying the expectation maximization (EM) algorithm, we are able to estimate signal and interference power using not only the pilots bits also the information bits. Since the information bits have smaller spreading factors, the estimator can have more samples. Moreover, in order to track the fasting fading, the number of samples used is small and the SIR estimate sequence is then noisy. We then apply a Kalman filter to further processing the EM-estimated sequence. The simulation results show that the proposed method can provide fast and accurate SIR estimates.