Pedestrian dead reckoning (PDR) is a navigation technique that provides and maintains the geographical position for a person travelling on foot by using self-contained sensors. In PDR techniques, each new position estimate is based on the previous estimate of the last step taking advantage of the sequential nature of pedestrian motion. In general, a PDR algorithm is composed of three parts: step detection, estimation of walking distance and tracking of sensor attitude. Most of PDR approaches consider low cost MEMS accelerometers, gyroscopes and magnetometers as the source of information. However, these sensors are affected by sensor noise and drift, which introduce errors in the displacement and relative attitude changes in the sensor´s frame of reference with respect to the human body. In order to improve the accuracy of the attitude estimation reducing the time-varying drift, this work presents the development of a Kalman Filter with Neuro-Fuzzy adaptation (KF-NFA), relying on information derived from triaxial accelerometer and gyroscope sensors contained in an inertial measurement unit (IMU). The adaptation process is performed on the filter statistical information matrix R, which is tuned using an Adaptive Neuro Fuzzy Inference System (ANFIS) based on the filter innovation sequence through a covariance-matching technique. Besides, in order to improve the distance traveled by the pedestrian, and consequently the localization accuracy, different types of activities are classified using a Multi-layer Perceptron (MLP) Neural Network (NN) according to extracted features based on Wavelet Decomposition. Basic activities that a pedestrian performs in his daily life, such as walking, walking fast, jogging and running, are considered. Subsequently the step-length is dynamically estimated using a multiple-input-single-output (MISO) Fuzzy Inference System (FIS). Validation testing and obtained results are presented.