H2020Докторантска мрежа2020–2025

TEVI · Time Encoded Voice Interfaces

„Хоризонт 2020“ — Действия „Мария Склодовска-Кюри“

Период: 2020-11-01 → 2025-04-30
Финансиране от ЕС: 752 715 €
Участници: 2
Схема: MSCA-ITN

Линиите свързват координатора с партньорите.

Накратко на български

Интегрирани микрофони с изкуствен интелект се разработват за по-ефективно разпознаване на глас и конкретни думи. Това помага за намаляване на консумацията на енергия в носимите устройства, за да работят те по-дълго с батерия.

Този кратък обзор е генериран от изкуствен интелект

Кратко обяснение, генерирано от езиков модел по текста на CORDIS. Оригиналът е по-долу.

Резултати накратко

Time Encoded Voice Interfaces

The development of microelectronics has enabled the deploment of artificial intelligence in portable and wearable devices. However, limiting the power consumption is crucial to enable battery operation. This requires the design of ultra-low power circuits, in the range of hundreds of nWs. Regarding smart portable applications, speech recognition is a main topic, in special voice activity detectors (VADs). VADs continuously detect human voice in noisy environments to trigger another more complex task like full-audio conversion or keyword spotting. This EID project focuses on MEMS integrated microphones incorporating artificial intelligence (edge computing). The conventional approach to VAD or Keyword Spotting digitizes first the input audio signal. Them Intensive digital computing is performed to either detect human voice or specific words. Although the feasibility of these architectures has been proven, a high power consumption is required, making them unsuitable for smart MEMS microphones that operate in devices with limited battery life. Time-encoded based solutions are proposed in this research project to overcome this limitation. The research project is divided into three different work packages, each one devoted to different processing stages, but with the major goal of being compatible towards a complete intelligent system: 1. Integration of MEMS-based microphones into the digitization stage. Direct encoding of the MEMS microphone with a pulse frequency modulated signal avoids the data conversion stage prior to VAD or keyword spotting. 2. Time-encoded feature extraction. Simplification of the conventional architecture composed of a set of band-pass filters and power estimators with ring-oscillator based filters for reduction of the silicon area and direct interface with the MEMS microphone. 3. Classification tasks using neural networks based on ring oscillators. Use of neuromorphic circuits that directly connect the neural network to the sensor, avoid pre-processing and feature extraction stages or rely on the outputs of workpackages 1 and 2.

Текст от CORDIS, на английски · Данни: CORDIS, © Европейски съюз

Цел на проекта

The research of this EID focuses on ultra-low power sensors incorporating artificial intelligence. The current solution for such systems requires an analog-to-digital converter (ADC) prior to the signal processing block, usually implemented with a neural network (NN). The innovation consists of removing the ADC prior to the NN by directly coupling the sensor to it and encoding the sensor signals with a voltage-controlled-oscillator (VCO). VCO-based ADCs have been used to implement integrated sensors. Achieving this goal requires to develop a new multiply-accumulate cell (MAC) for the first layer of the NN that operates with signals from the VCO, and a suitable VCO interfaces with existing sensors. In most applications, the raw data form the sensor is required as well. Here, signals coming from the VCO can also be converted to a sampled sequence by enabling a digital decoder, which is not needed when detecting a pattern in the NN. As a benefit, power consumption can meet the requirements of battery-operated products. Power improvement comes from both the ADC removal and the power efficiency of the NN implementation. Approaches to implement a sensor interface using a VCO and to implement a phase/frequency-encoded MAC unit (P-MAC) for a NN have been attempted separately but, there is no combination of both ideas. The research in this EID tries to bridge this gap. This architecture can be useful for both research and industrial applications, such as neural probe chips, wearable electronics or battery powered IoT devices. This EID proposal requires intersectoral involvement of both academia and the industry, to develop a doctoral program and train researchers that will be in high demand by having the specific skills developed in this research. We have selected waterproof smart microphones as an application to benefit from this research, which may directly lead to a product development of interest to microelectronic industries in the EU producing MEMS microphones.

Оригинален текст от CORDIS (на английски).

Участници

UNIVERSIDAD CARLOS III DE MADRID · Getafe (Madrid)КоординаторИспания
INFINEON TECHNOLOGIES AUSTRIA AG · VillachАвстрия