Columbia Technology Ventures

Compact, Ultra-Low-Power Time-Delay Estimation Circuit for Sound Localization

This technology is a time-delay estimation circuit to localize the source of a sound in a 3D space and boasts 10,000-fold less power consumption, simpler architecture, and a smaller form factor than current technologies.

Unmet Need: Low-power time-delay estimation for sound localization

Localizing a sound in a complex environment has found many uses such as in sonar, hearing aids, speaker localization, navigation, and surveillance. As IoT and personal electronic devices become increasingly intelligent and aware of the environment, power consumption of various sensors is a critical constraint for these systems. State-of-the-art time-delay estimation requires high power consumption due to complex time-delay estimation algorithms, high number of analog-to-digital converter (ADC) bits, or requirement of large memory blocks. To lower power consumption of these devices, there is a need for simpler circuit architectures and advanced time-delay estimation algorithms.

The Technology: Polarity-Coincidence Correlation Adaptive Time-Delay Estimation – PCC-ATDE

This technology describes a time-delay-to-digital converter that achieves power-efficient analog-digital mixed signal processing and is capable of estimating the time-delay of multiple microphone input signals. This circuit utilizes feedback-based polarity-coincidence correlation (PCC) time-delay estimation, which requires less memory, offers a faster convergence rate, and boasts higher stability than generalized cross-correlation function (GCC)-based algorithms. It provides simpler architecture than the current state-of-the-art solutions as it does not require multibit ADCs, large memory blocks, or complex arithmetical operations. As such, this technology can greatly enhance the energy-efficiency of IoT and portable electronic devices.

A proof-of-concept application of PCC-ATDE has been implemented in a wearable headset platform that could accurately estimate the distance of an approaching vehicle in a busy street setting.

Applications:

  • Sound localization in low power devices such as IoT, personal electronics, and hearing aids
  • Automotive awareness system
  • Traffic control
  • Pedestrian safety
  • Drone localization
  • Remote pipeline control
  • Detection of machinery and sensor malfunction
  • Wearable, implantable biosensors

Advantages:

  • Up to 10,000-fold more power efficient than state-of-the-art technologies
  • Ultra-low power utilization of <100nW
  • Allows for always-on operation with limited battery power
  • 1-bit quantizer requires only simple arithmetic processing
  • Does not require large memory blocks
  • Small footprint requires few components

Lab Director:

Peter Kinget, Ph.D.

Patent Information:

Patent Issued

Related Publications:

Tech Ventures Reference: