Study On Digitalization Of Student Terminal In Language Learning System

Language learning system is an important multimedia teaching equipment, which can help students improve their ability of speaking and listening foreign languages. It is a practical application of speech processing technology in multimedia teaching. Most of the related products are analogue systems at home and abroad, which have many shortages. Therefore, the development of digital language learning system has becomg an urgent need.Student terminal is one of the key units in language learning system, whose digitalization is very important to the whole system. It consists of speech processing, man-machine interactive channel and transmission interface. Speech processing is the kernel of student terminal, whose main function is speech acquisition, speech compression/decompression and broadcast of audio files. Man-machine interactive channel gives students a good interactive interface, including liquid crystal display and keyboard. The function of transmission interface is data transmission between student terminal and sub-exchange.However, the digital speech needs much higher code rate than analogue signals resulting in high pressure on data transmission. Therefore, it is necessary for us to have speech compression coded and make it a key part in this paper. Our research object is Conjugate Structure Algebraic Code Excited Linear Prediction (CS-ACELP) algorithm with its good performance, whose synthesized speech quality at 8kbps is not worse than that of ADPCM at 32kbps.The main contents in this paper are as follows:①According to the principle of speech signal generation,this paper introduces the digital model of speech signal generation. It also studies the key technology in speech signal processing, such as short-time analysis, linear predictive analysis,vector quantization and Code Excited Linear Prediction , which lays the theoretical foundation for analyzing principle of CS-ACELP algorithm.②The paper reviews the history of speech compression coding, and then introduces the principle of CS-ACELP algorithm in details.③It is very difficult in DSP to realize CS-ACELP algorithm in real-time because of its large computation. The computation of LSP coefficient computation, pitch detection, codebook research consumes about 70%～75% of the whole. As a result, this paper makes a study of the improvement of pitch detection,adaptive codebook research and fixed codebook research as well as the speech endpoint detection and sample rate conversion, which are all simulated with MATLAB.④The author has proposed the system scheme and the soft hardware design of student terminals. At last, the coded/decoded processing of speech signals is realized in real-time with CS-ACELP algorithm in TMS320vc5402.