Office of Technology Transfer – University of Michigan

Ultra Low Power Speech Processor using Memristor Crossbar Memory

Technology #5931

The size and weight of hearing aids and cochlear implants is significantly determined by the battery, which is chosen according to the power usage of the device. High power usage requires a larger, heavier battery, which can increase discomfort and draw unwanted attention to the device and the user. University of Michigan researchers are attempting to alleviate this problem through a new technology that uses memristor crossbar memory to significantly decrease the power use in the speech processor, traditionally one of the most power-hungry components in a hearing-assistive device.

Information retained, power saved during memory shutdown

Memristor crossbar memory can be densely packed and is non-volatile, meaning it retains information stored one it even when the memory is powered off. The new speech processor design takes advantage of this feature and disables power to the memory when access is not required during each operation. Simulations of the speech processor demonstrate almost 80% power savings per processor operation compared to existing low-power speech processor standards, without loss of data. The lower power usage, along with improved memory density, can help reduce the size and weight of the battery, memory chip, and overall device, making it more comfortable to wear, longer lasting, and discreet.

Applications

  • Cochlear implants
  • Digital hearing aids

Advantages

  • Significantly reduced power usage compared to current volatile memory speech processors
  • Smaller size due to density gains of memristor crossbars