6C-05
Voiceprint Recognition Using Adaptive Mel-Frequency Feature Extraction
○包 立澳,竹縄知之(東京海洋大),左 毅(大連海事大)
This study presents a voiceprint recognition framework based on an adaptive feature extraction method that improves the traditional Mel scale transformation. By replacing fixed Mel-scale filters with adaptive ones, the method adjusts frequency parameters to better capture important speech features. These features are then classified using a ResCNN-Attention Network, enabling reliable recognition in different environments. The adaptive transformation makes feature representation more effective, showing promise for advancing voiceprint recognition.