Intel® Natural Language Processing Won’t Hit the Wall Anytime Soon

Delivering language and machine learning capabilities to eHealth—with an eye toward automotive, robotics, and more

There is a transition in healthcare, whereby electronic equipment and software are no longer used by trained staff to identify, diagnose and record a patient’s health; now devices and software are in the hands of the patient for management of health and wellbeing.

Radar Pace typifies the challenges of the eHealth market, namely to be lightweight, power-sensitive and at an attractive price point.

The definition of eHealth can be all encompassing. Essentially it is the use of information and communications technologies in healthcare. Within that definition however, there is the management of medical records, enhanced imaging and surgery training, and online health portals for diagnosis. This article focuses on the area of health surveillance, where patients manage and analyze their own health.

The American Health Information Community (AHIC) issued recommendations to develop eHealth designed to increase awareness levels for mobile applications and wearable, remote monitoring devices. Grand View Research cites management and monitoring as one of the major drivers of the eHealth market. It estimates that the market will be worth $308 billion by 2022.

Wearable Weight
One of the biggest challenges in designing wearable eHealth devices is that they have to be just that —wearable. Not only do they have to be lightweight, they have to be almost ‘invisible,’ so they are not cumbersome during activities where they collect data, such as running on a treadmill. Considerable excitement for wearable devices arrived last month with the introduction of Oakley’s Radar Pace eyewear, designed and manufactured by eyewear specialist, Luxottica Group, and using Intel® Real Speech technology. The real-time, voice-activated coaching system brings the voice of a coach to users via eyewear.

Figure 1: [photo courtesy Oakley]

Figure 1: (photo courtesy Oakley)

First introduced at CES 2016, the eyewear brings language and machine learning capabilities to eHealth. Intel® Real Speech technology is built on complex language and reasoning models, allowing Radar Pace to interpret tens of thousands of phrases. Runners short on breath after completing a run do not have to remember the ‘correct’ terminology to ask, “Are we done yet?” To achieve this, Intel has had to combine grammar and statistical language models with machine learning technologies. eHealth is far from the only sector for which embedded designers can now think about incorporating natural language and hands-free interaction. Intel points out that the attributes of ithis breakthrough natural language processing technology apply as well to the navigation and entertainment systems for our vehicles, in our homes, including in the form of personal assistants, and within the realm of industrial robotics.

Figure 2: Oakley’s Radar Pace eyewear relies on Intel’s Real Speech technology for contextual, natural-sounding two-way coaching [photo courtesy Oakley, Inc.].

Figure 2: Oakley’s Radar Pace eyewear relies on Intel’s Real Speech technology for contextual, natural-sounding two-way coaching (photo courtesy Oakley, Inc.).

Using speech recognition, natural language understanding and generation, language-based reasoning, and Artificial Intelligence (AI), the wearer can initiate or respond to meaningful interaction. In addition, the ability to enable context via in-app analytics, means that future conversations can be meaningful; Radar Pace can suggest a workout routine, for example, based on the wearer’s known ability or thresholds. In addition to assessing the variations of voice patterns and whether to use a male or female voice, all of this had to be packed into a 56g eyewear frame.

Radar Pace typifies the challenges of the eHealth market, namely to be lightweight, power-sensitive, and at an attractive price point. Analyst IDC reports that while the eyewear category will take up less than 10% of wearable device shipment pies by 2020, it will nevertheless be able to take credit for more than 40% of “the total revenue of the Wearables market due to the high prices for specialized commercial devices.”

New Audiences

At this month’s electronica exhibition in Munich, Germany, visitors saw a health patch that can track physical and cardiac activity while monitoring bioelectrical impedance. The Holst Centre (a collaboration with research institute, imec and the independent research group, TNO) introduced the health patch (Figure 1) which is available for licensing by partner companies.

Figure 3: The health patch was developed by the Holst Centre at its High Tech Campus in Eindhoven, the Netherlands.

Figure 3: The health patch was developed by the Holst Centre at its High Tech Campus in Eindhoven, the Netherlands.

The patch has been tested in a controlled environment and is ready for pre-clinical and usability studies to track and manage the wearer’s health. Ruben de Francisco, Program Manager Wearable Health at imec and Holst Centre believes it has enormous potential: “Looking ahead, we plan . . . to lay the foundation for a powerful patient management solution that not only captures data, but that also turns data into meaningful information upon which people and health providers can act.”The small form factor patch relies on low power consumption. It combines a power-optimized chip with sensor functions such as an accelerometer, ECG tracking and bioelectrical impedance monitoring to measure tissue, respiratory activity, in the same electrode patch. It relies on Shinko Electric Industries’ System in Package (SiP) miniaturization technology and batteries for wearables from Hitachi Maxell.

The month before the electronics fair in German, the Holst Centre also announced a sensor hub System on Chip (SoC) which combines biomedical analog interfaces, on-board Digital Signal Processing (DSP) and multi-day monitoring using a single battery (Figure 4).


Figure 4: The biomedical sensor hub SoC has been developed by the Holst Centre, specifically for wearable health devices.

There are three ECG channel analog interfaces, together with photo-plethysmography (PPG), galvanic skin response (GSR), and two multi-frequency bio-impedance (BIO-Z) channels, targeting applications such as impedance-tomography, body fluid analysis and stroke volume measurements. There are also three reconfigurable channels. The SoC can therefore be used in patch monitors, chest-band heart rate monitors, respiration or hydration monitors and devices to calculate blood pressure.

hayes_caroline_115Caroline Hayes has been a journalist, covering the electronics sector for over 20 years. She has worked on many titles, most recently the pan-European magazine, EPN.

Share and Enjoy:
  • Digg
  • Sphinn
  • Facebook
  • Mixx
  • Google
  • TwitThis
Extension Media websites place cookies on your device to give you the best user experience. By using our websites, you agree to placement of these cookies and to our Privacy Policy. Please click here to accept.