Best foot forward for gait recognition

Materials World magazine
,
3 Jul 2017

Not only could CSIRO’s new wearable technology generate its power from footsteps – it could use a wearer’s gait for identification, too. Simon Frost reports.

Wearable devices are a growing market, with health monitoring technologies such as Fitbit and Apple Watch leading the way. In its Wearable Technology 2017–2027: Markets, Players, Forecasts report, technology market research group IDTechEX predicted that the market for personable wearable technologies will reach US$150 billion by 2026. But how can we protect the personal data they collect, and how best to power them? Scientists at Data61, the digital arm of Australian government research agency CSIRO, have come up with a new system that combines kinetic energy harvesting (KEH) with gait recognition to address both questions.

KEH-Gait uses the output voltage signal of a KEH device as its source of gait recognition, as researcher Sara Khalifa explained to Materials World. ‘We used a piezoelectric cantilevered beam to implement a KEH transducer. When the vibrations generated from human motion subject the piezoelectric material to mechanical stress, it expands on one side and contracts on the other,’ she said. 

‘Positive charges accumulate on the expanded side and negative charges on the contracted side, generating an AC voltage as the beam oscillates around the neutral position.’ The voltage is proportional to the applied stress, meaning that different people generate a different AC voltage depending on their walking style. 

The researchers took interest in authentication via gait recognition because it is both an easy way to submit data and a difficult signature to mimic. ‘As we walk around each day our gait can be sampled continuously and verified without us having to manually adjust anything,’ Khalifa said. ‘It is also more secure than a password […] the KEH-Gait keeps authenticating the user continuously, so it collects a significant amount of information about our movements, making it difficult to imitate or hack.’ 

To achieve an accuracy of 95%, the KEH-Gait obtains five ‘gait cycles’, each comprising two successive steps. The output voltages of the KEH are subject to pattern recognition techniques – signal processing and classification – to authenticate the user. 

‘First, the gait cycles are segmented from time series voltage signal and then interpolated into the same length. We then apply some signal processing techniques to remove noise and outliers of gait cycles. The obtained gait cycles are used in a classification process, employing multi-step sparse representation classification, which efficiently fuses information from multiple steps,’ Khalifa explained. 

Authentication is likely to become increasingly important in medical applications, as data from such devices could be used to inform health insurance premiums or renew prescriptions. ‘Before transmission to the data centre of a healthcare company, the device first collects gait data and transmits them to the server. The server would then verify the user’s identity using their gait data. If the user passes authentification, the private data such as blood pressure or heart rate would then be transmitted to the server, while a failed verification would prevent the data from being transmitted,’ Khalifa said. 

The authentication aspect is the focus of the team’s recent work, but using KEH to help or entirely power wearable devices would also be useful. Khalifa explained, ‘The generated AC voltage should be rectified to produce DC, which can be used to power or improve the battery life of wearables and mobile devices.’

The team is now developing a prototype that is capable of harvesting a greater amount of power from a user’s movement, which as well as potentially leading towards self-powered wearables, would improve the accuracy of the technology’s authentication. 

Current KEH materials provide one-dimensional energy generation patterns, and Khalifa notes that materials able to harvest energy along three axes would improve both energy generation and data potential, the ultimate aim being a device that powers itself entirely and offers 100% accurate authentication, as multiple sources of information linked to movement could make a motional signature impossible to mimic.  

Aside from medical monitoring, Khalifa notes that the KEH-Gait technology could be used to unlock cars, bank terminals or even verify a passport holder’s identity. Research continues, one step at a time.