Recently, privacy concerns have increased in public and office speech applications. Parametric array loudspeakers (PALs) achieve sharp directivity by emitting intense ultrasonic waves. The ultrasound is modulated with an audio signal, and the target sound is reproduced as a difference frequency due to the nonlinearity of air. However, speech leakage can still occur along the radiation axis. A conventional method uses separate emission of the carrier and sideband waves from different PALs, but it suffers from intermodulation distortion that can cause leakage resembling the original speech. Our previous method reduced leakage by dividing the sideband into multiple frequency sub-bands and assigning them to separate PALs. However, it required more PALs and lacked flexibility in designing the frequency partitions. To address this, we propose sideband decomposition in double sideband modulation with suppressed carrier (DSB-SC), dividing the sidebands into multiple frequency sub-bands that are then assigned to two different PALs. The proposed method achieves speech leakage reduction by assigning different frequency sub-bands to multiple PALs. Thus, the leaked sound becomes less intelligible outside the audio spot compared to conventional separate emission method. We evaluate the effectiveness of the proposed method based on speech intelligibility both inside and outside the audio spot.
Iwagami et al. (Wed,) studied this question.