Speech visualization based on improved spectrum for deaf children

A method of computer analysis and display which represents the vowel is described. Efforts to date have been devoted to speech feature extraction and speech display. Speech features are extracted by the improved spectrum method, which combines Fourier Transform with auditory characteristics. A novel speech plot is proposed. Firstly, speech signal undergoes a series of preprocessing course. Secondly, we process spectrum analysis for speech signal by the proposed method. The spectral energy value of all sample points are graded by 4 sections. We can calculate the sum of all frequency sample point spectrum energy at one time .A new feature vector is produced to represent speech signal, which is encoded to be displayed by plot patterns. Experiments show this has more applications in computer-based speech training aids and the deaf children can make use of their own brain to identify speech for training their oral ability effectively.