Based Be Sine + Noise Model Representation Of Wideband Audio Signals And Compression

The development of Communications and Digital Signal Processing technology in both software and hardware has provided more multimedia applications. The signal modeling and compression have been the hot topics in both academic and industry research areas. Audio coding algorithms with such features as low bit-rate, high fidelity and real-time applicable are reported to have prosperous future. Starting with research on audio signal modeling, this thesis presents a signal modeling technique which divides general audio signal into two parts: Deterministic Part and Stochastic Part, i.e. a Sinusoidal plus Noise (S+N) modeling of audio signal. Improvements to the classical S+N model are also introduced. Then the thesis presents the fundamentals of Perceptual Audio Coding and its core technique: Psychoacoustic Model. Based on the signal modeling theory and psychoacoustic theory, a low bit-rate wide-band audio coding scheme is proposed. From the signal modeling point of view research on S+N model, which has been widely used in Speech Coding, has been carried out. Sinusoidal Model based on both Analysis-Synthesis method and Analysis-by-Synthesis (AbS) method has been implemented and compared. According to characters of wide-band audio signal, improvements to classical sinusoidal model has been proposed. With research on Perceptual Audio Coding theory, Psychoacoustic Model described in MPEG(Moving Pictures Experts Group) Audio part has been implemented and integrated to sinusoidal parameter picking and quantization. To give further bit reduction of the redundancy between signal frames, a novel Sinusoidal Tracking Coding algorithm has been proposed. Frequency domain phaseless reconstruction is therefore implemented based on sinusoidal tracking. The general bit-rate of 0.5 bit per sample is achieved and the bit-rate can be scalable from 6kbps to 16kbps. For 32kHz sampled, 16 bits quantized linear PCM signal, the coding scheme proposed in this thesis can provide reconstructed audio with quality from reasonable to nearly transparent. Comparison between this codec and the state-of-the-art audio coders shows that it gives better quality and wider bandwidth at the same bit-rate; it need less bit-rate to provide similar quality.