Scalabled Multi-rate Speech Coding Algorithm Based On ILBC

while the speech signal transmission in the network, The situation such as the channel load increase, the bandwidth reduction, are easy prone to packet loss, jitter, latency ect, These reduces the voice quality. So it is important to improve the codec quality to code the speech. The main purpose of the speech coding is on the premise of maintain the algorithm complexity and communication time delay, Take as little as possible communication capacity and transfer as far as possible high quality speech. In order to in the process of transmission errors with high robustness and ensure the quality of speech, lowing code rate, save bandwidth resources, Many speech decoding algorithm is proposed, such as G.729、AMR、EVRC and iLBC, As well as the improvements on these algorithms. Which iLBC(Internet Low Bit codec) uses a linear predictive coding technology based on the frame, Compared with other decoding algorithm, The encoder has a dynamic code update technology, Packet loss hiding technology and speech enhancement technology. This speech codec based on the IP packet switching network has very good performance, When network environment is poor can still provide high quality speech service. However, iLBC processing the data processing rate is not high flexibility, When dealing with packet loss, compared with code excitation linear prediction model of the codec, it would take a higher bit rate, This increased the network load, cause the delay. Aiming at these problems, This paper studies the extension multi-rate speech codec algorithm based on iLBC. The main work is as follows:This paper introduces the principle of some extension encoder in detail, Analyzed the quality of the reconstructed speech coding algorithm. And then expounds the basic process of iLBC, Then improve on the standard of iLBC encoder, First of all, according to high energy compression features of the discrete cosine transform, requires less number of bits can describe the characteristics of the initial state. So use the discrete cosine transform to the initial state of speech signal instead of the time domain on scalar quantization. Then in order to slow down the sending end encoder for speech signal coding appear lost package the effects on the voice quality. And increased feedback adjustment module in the previous steps, real-time adjustment of speech quality. Finally, in order to make the encoder can support high quality audio signal, increased broadband coding module on the basis of narrow-band, This method is mainly based on the layer bitstream processing, And the narrowband and broadband coding process are described in detail. Through the simulation experiment, comparing with the classic speech codec algorithm, Simulation their performance under different packet loss rate and code rate, After processing of speech signals the algorithm in this paper is more natural and can understand sex is higher also.