| 摘要: | 類神經網路是利用人工神經元模擬生物神經系統的一種預測演算法,透過類 神經網路可以進行大量資料的平行運算及預測。類神經網路的發展是為了模擬生 物的聽覺及視覺系統,所以在於影像識別及語音訊號部分的處理效果都特別出色。 由於類神經網路可以實作於硬體上,而且具有平行運算以及一般化的能力,所以 可以運用在各種不同領域範圍上並進行高效能的預測分類運算。
訊號處理的範圍廣泛,諸如圖像、視訊以及語音訊號都有不同的訊號處理方 式。線性預測編碼是一種於時間域上的線性預估模型演算方式,線性預測編碼可 以將一連串的輸入訊號轉換成一線性組合方程式,藉由線性預測編碼取出輸入訊 號的線性預測係數,並做為該輸入訊號的特徵值,進而進行預測或者運用於訊號 壓縮的技術上。
由於自然界的各種訊號可以由各種訊號波組成,諸如正弦波、餘弦波、方波、 三角波等,線性預測編碼於未知訊號的特徵取樣上有良好的表現,透過線性預測 編碼取得未知訊號的線性預測係數做為該訊號的特徵值。類神經網路的訓練學習 及預測有良好的表現,而且可以運用於各種輸入樣本上。本論文所自製的訊號為 常見的訊號所組成,例如:方波、正弦波、三角波等,使用本論文所產生的自製 訊號組合進行線性預測編碼的運算,並透過類神經網路對經過線性預測編碼取得 的特徵值做訓練學習以及分類預測。本論文最後的結果呈現使用本論文所提出的 機制於第一名的預測結果為100%,證明本機制於分類預測上能提供良好及準確 的表現。
Artificial Neural Networks are predictive algorithms which use artificial neurons to emulate biological nervous system, and it can provide parallel computing for large scale data. Neural networks are originally used and developed for biological vision and auditory systems, so it has excellent performance in speech and image recognition. Because neural networks can be implemented on hardware and it has the ability of parallel computing and generation, so neural networks can provide high performance in data prediction and classification for other domain.
Signal Processing is widely used in analyzing image, video and speech. Linear Predictive Coding is an algorithm of linear predictive model in time domain. It can convert continuous signals to a linear combination equation, which is used in prediction or signals compression by fetching the Linear Predictive Coefficients of input signal. In this paper, the authors propose a mechanism that uses Linear Predictive Coding (LPC) to fetch the feature values of multi-source input signals, and then use neural networks to deal with the feature values. To emulate natural signals, the source signals are made and combined with well-known waves, such as Sine, Cosine, and triangle wave. The simulation results show that the proposed mechanism can provide 100% accuracy of correct prediction on highest similarity, which proves that proposed mechanism can provide great accuracy in prediction. |
