Using GPS-derived speed patterns for recognition of transport modes in adults

Abstract

BACKGROUND: Identification of active or sedentary modes of transport is of relevance for studies assessing physical activity or addressing exposure assessment. We assessed in a proof-of-principle study if speed as logged by GPSs could be used to identify modes of transport (walking, bicycling, and motorized transport: car, bus or train). METHODS: 12 persons commuting to work walking, bicycling or with motorized transport carried GPSs for two commutes and recorded their mode of transport. We evaluated seven speed metrics: mean, 95th percentile of speed, standard deviation of the mean, rate-of-change, standardized-rate-of-change, acceleration and deceleration. We assessed which speed metric would best identify the transport mode using discriminant analyses. We applied cross validation and calculated agreement (Cohen's Kappa) between actual and derived modes of transport. RESULTS: Mode of transport was reliably classified whenever a person used a mode of transport for longer than one minute. Best results were observed when using the 95th percentile of speed, acceleration and deceleration (kappa 0.73). When we combined all motorized traffic into one category, kappa increased to 0.95. CONCLUSIONS: GPS-measured speed enable the identification of modes of transport. Given the current low costs of GPS devices and the built-in capacity of GPS tracking in most smartphones, the use of such devices in large epidemiological studies may facilitate the assessment of physical activity related to transport modes, or improve exposure assessment using automated travel mode detection.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 30 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.