Hybrid Speech Compression method Based on DWT and SVD

摘要

Speech compression attractive topic related to human communication so it is necessary to update methods to find efficient and low bit rate method. The main issue is keep speech quality. Methodology present compression method consist of sequences of steps; reemphasis speech using high pass filter as preprocessing followed by voice activity detection to remove silence interval. Framing a result to specific number of samples, each frame will reordered into two dimension matrix. It transformed with discreet wavelet transform (four component). Three of them analyzed using single value decomposition and get parts of them with the remaining part to reconstruct frame. Huffman codding applied to selected part to reduced bitrate of data. The result shows that compression factor reach about 6 to 9 rate with acceptable quality.