Detecting outdoor walking workouts on a wearable device

A method for improving performance of a wearable device while recording a walking workout, the method comprising:

• training, by a processor, a predictive model to detect start and end points of a walking workout, the training based on training data that includes motion data indicative of casual walking activities and walking workouts;
• measuring, by a motion sensing module of the wearable device, motion data of a user;
• determining, by the processor circuit of the wearable device, a number of steps performed by the user based on the motion data;
• detecting, by the processor circuit, a bout based on the number of steps performed by the user during a predetermined period of time, the bout including a plurality of continuous steps performed by the user;
• detecting, by the processor circuit, a step rate based on the number of steps performed by the user during the predetermined period of time;
• determining, by the processor circuit, mechanical work performed by the user during the bout, wherein the mechanical work is based on a pedestrian work model that is a function of at least the step rate and a load;
• detecting, by the processor circuit, a start of the walking workout based on the mechanical work, wherein detecting the start of the walking workout includes:
• comparing the mechanical work performed by the user to a mechanical work threshold;
• detecting a value for the mechanical work performed by the user that is below the mechanical work threshold;
• in response to detecting the value for the mechanical work performed by the user that is below the mechanical work threshold, determining, with the trained predictive model, a patterned movement indicative of walking based on the motion data;
• in response to determining the patterned movement indicative of walking, starting the walking workout and
• recording, by the processor circuit, the walking workout.