| Popis: |
Ultrasound based tracking and localization are more and more popular in recent years. However, to achieve 3D tracking or localization, at least three sound sources are needed. Unfortunately, there are only two sound sources available in most scenarios in daily life. To address this problem, we propose SOLO (Single sound sOurce LOcalization with single microphone), a novel approach to infer 2D information from single sound source. We found that sound with different frequencies from a signal source may have different strength distribtuion in different areas. Each area has a unique group of sound strengths of different frequencies. So we can take this question as a classify question. Based on this, we first obtain the signal strength in different frequencies by using STFT (Short-Time Fourier Transform). Next we measure 1D distance with traditional phase-based distance measuring approach. Then we take 1D distance and the sound strengths as features to a neural network to classify the sample into different classes, each representing a region in 2D space. With our system, we can distinguish 3 × 3 regions (5 cm × 5 cm for each region) with only single sound source and single microphone. SOLO is the first work to achieve 2D localization and tracking with only single sound source and single microphone as far as we know. This work make it possible to realize 3D tracking or localization with only two sound sources. |
