Fast Sound Source Localization Based on SRP-PHAT Using Density Peaks Clustering

Sound source localization has been increasingly used recently. Among the existing techniques of sound localization, steered response power–phase transform (SRP-PHAT) exhibits considerable advantages regarding anti-noise and anti-reverberation. When applied in real-time situations, however, heavy computational load makes it impossible to localize a reasonable time since SRP-PHAT employs grid search scheme. To solve problem, an improved procedure called ODB-SRP-PHAT, i.e., power phase transformation with offline database (ODB), was proposed by authors. The basic idea ODB-SRP-PHAT is determine possible positions using density peak clustering before store identified ODB. Then, at online positioning stage, only values ODB will be calculated. monitoring, e.g., locating speaker video conference, significantly smaller than that SRP-PHAT. Simulations experiments under real environment verified high accuracy small ODB-SRP-PHAT. In addition, anti-reverberation remained. suggested displayed good applicability environment.