Research Article
Data-Driven Decision-Support System for Speaker Identification Using E-Vector System
Algorithm 1
E-vector system for speaker identification.
Input: z (continuous speech signal); frame 20 ms; Step 10 ms; n = X | Output: E-vector ; accuracy; micro-F1 | (1) | Initialization: c (a)⟶IDCT cepstrum coefficient | (2) | Data preprocessing: c (a) = data preprocessing (z) | (3) | Cluster: Set S (S←), Name c (a)⟶ | (4) | For each in A do | (5) | | (6) | Put all into S | (7) | If | (8) | | (9) | If n = X | (10) | break; | (11) | else | (12) | continue; | (13) | end | (14) | end | (15) | end | (16) | Learn models: put A into GMM, HMM ⟶ accuracy, micro-F1 |
|