Pitch shifting techniques for high-frequency passive sonar audio

Abstract

Listening to passive sonar signals is a vital tool for sonar operators to classify underwater sound sources. While many passive sonar systems operate in the human auditory range (20 Hz to 20 kHz) there are a considerable number of high-frequency systems that extend beyond this range. This report examines pitch shifting algorithms for compressing ultrasonic, bandlimited passive sonar signals down into the auditory spectrum. By utilizing pitch shifting techniques the signal’s harmonic structure and length in time are retained. The frequency spectrum is lowered into the auditory range so that the sonar operator may then listen and characterize targets. Three pitch shifting algorithms are examined: Waveform Similarity Overlap-Add (WSOLA), Phase Vocoder, and Constant-Q Transform (CQT). Both synthetic and real sonar data is experimentally applied to each method and results are presented. Comparisons of performance are provided with an emphasis on feasibility for real-time sonar system implementation.