Wearable device has been one of the most popular consumer products these days. One of the key features for wearable device is the sport and health application. For example, users get used to record the running trajectory and odometer by using the wearable GNSS navigation system and track their sport performance from the record. Nowadays, wearable device supplier utilizes the micro-electro-mechanical systems (MEMS) inertial measurement unit (IMU) to further expand the applications, such as activity recognition and pedometer. However, these two technologies have not been tightly cooperated. Due to the size and hardware limitation of the antenna design in wearable device, GNSS performance is fragile in GNSS-harsh environments, urban area, forest canopy. Especially for the speed information, it is vulnerable to unstable pseudo range and doppler measurements from bad satellites. Moreover, bad speed estimation will also result in the performance of trajectory and odometer, which are the two major standards for the users to evaluate their sport performance.

Airoha Technology (A subsidiary of MediaTek) proposes FDR, the novel tightly integration with the smart sensor and GNSS engine together to overcome the trajectory and speed issues in wearable applications. FDR (Fitness Dead Reckoning) refers to GNSS data fused with IMU data in fitness and outdoor activities. There are two innovation steps, smart sensor deployment and tightly integration with GNSS.

1) In smart sensor application, the proposed fitness motion detection is developed. There are three states to recognize, motion, pose and movement states. Firstly, the motion states are defined as sport activities, such as walking, jogging, and running. Secondly, pose states are defined as device pattern, for example, swing and look-up states. Finally, the movement states present the information of static, forward, backward, etc. To well detect those human activities, 6-axis IMU sensor data is fully applied, including the time-domain (standard deviation, mean, maximum, minimum, etc.) and frequency domain (activity, mobility, complexity) feature extraction.

2) To improve GNSS performance, the smart sensor aiding is developed for different scenarios and fitting for different sport activities. In proposed tightly integration, each state is corresponding for different fusion strategies. The firs aiding is the measurement selection with smart sensor. The relationship of doppler measurement and pose states is evaluated and analyzed, especially for the arm-swing pattern. This information can remove the unstable measurement to improve the speed estimation of GNSS engine. To enhance the speed estimation for odometer application, the motion states are applied in GNSS engine. According to the motion states, we adapt the speed models for running, jogging, and walking to generate the accurate ground speed information as a measurement in GNSS engine.

We have designed several experiments to evaluate the proposed fusion strategy. First, the influence of arm-swing pattern has been analyzed corresponding to the doppler measurement in urban, forest canopy and open sky areas. Second, the watch platform and reference system are compared in GNSS-harsh environments. The results show that by applying the swing pattern, inaccurate satellite measurement can be selected and excluded in GNSS engine. Further, the speed and trajectory performance are both improved 15% by using the proposed smart sensor information, especially up to 30% for the environments where top 4 mean signal-to-noise ratio (SNR) is under 30dB.

Learn more on https://www.ion.org/gnss/abstracts.cfm?paperID=12430