Speech feature extraction using linear Chirplet transform and its applications

Most speech processing models begin with feature extraction and then pass the vector to primary model. The solution's performance mainly depends on quality of representation model architecture. Much research focuses designing robust deep network architecture ignoring representation's important role during neural era. This work aims exploit a new approach design signal in time-frequency domain via Linear Chirplet Transform (LCT). proposed method provides sensitive frequency change inside human solid mathematical foundation. is potential direction for many applications. experimental results show improvement based LCT compared MFCC or Fourier Transform. In both speaker gender recognition, dialect significantly improved other features. result also implies that independent language, so it can be used various