 Hello and welcome to this presentation about Embedded Advanced Pedometer. The new generation iNemo inertial measurement units embed an advanced pedometer directly on silicon. The algorithm runs in an ultra-low power domain, ensuring extensive battery lifetime in battery constrained applications. Thanks to its superior configurability, the advanced embedded pedometer is suitable for a large range of applications ranging from smartphones to wearable devices. The algorithm processes and analysis accelerometer waveform in order to count user steps during walking and running activities. ST freely provides the support and the tools for easily configuring the device and tuning the algorithm configuration for a best-in-class user experience. The algorithm is composed of a cascade of four stages. First of all, the signal magnitude is computed. In this way, the acceleration signal is independent from device orientation. The second stage consists in a fur filter able to extract relevant frequency components and to smooth the signal by cutting off high frequencies. Then as third step, a peak detector is able to find the maximum and minimum in the waveform and to compute the peak-to-peak value. In the last phase, if the peak-to-peak value is greater than a threshold, a step is counted. The first level of false positive rejection is implemented inside the algorithm core. The internal threshold for step detection is updated after each peak-to-peak evaluation. It is increased with a configurable speed if a step is detected or decreased with a configurable speed if a step is not detected. This approach ensures high accuracy when the user starts to walk and false peaks rejection when the user is walking or running. Moreover, the algorithm comes with an internal configurable debouncer able to filter false walks. Indeed, an accelerometer pattern is recognized as a walk or run only if a minimum number of steps are counted. The advanced pedometer comes with an additional false positive recognition block able to enhance the filtering of false steps counting. This algorithm performs real-time recognition of walking and running activities based on statistical data and inhibits the step counter if no walking activity is detected. As per the other pedometer blocks, it is fully reconfigurable. The pedometer works at 26 hertz pace and is not affected by the selected device power mode, such as ultra low power, low power, high performance, thus guaranteeing not only an ultra low power experience, but also extreme flexibility in conjunction with other device functionalities. As an example, the pedometer outputs could be batched in the device FIFO in order to decrease overall system current conception even more. Moreover, accelerometer operating mode can be changed at runtime and based on user requirements without impacting pedometer performance. The pedometer function requires an incredibly low current conception down to 1.5 microampere or 4.5 microampere, in the case of false positive recognition enabled. This is even more unbelievable considering its accuracy of 98.2%. ST offers to its users the possibility to tune the algorithm through Unico graphical user interface. The regression tool is a feature of Unico GUI, having the aim of automating the fine-tuning process of step counting capabilities based on a collected input dataset. It allows to generate an optimal configuration based on the collected input dataset through an iterative algorithm developed in the tool. In this way, it is not requested to the user to know about pedometer parameters and theory. Once regression is finished, the tool generates the device register configuration, which sets the internal pedometer parameters. The dataset, which is passed to the regression tool as input, is a folder containing acceleration pattern files. Each file should contain the accelerometer data acquired during the user walking or running, and the file must contain, in its file name, the effective number of steps as shown in the list. It is mandatory to log the data at 26 hertz output data rate in millig format with line breaks. Once the dataset is ready, it is time to run the regression. Connect st eval mki 109 v3 board to the pc through a usb cable and start Unico application. Then select inemo inertial modules on the left list view. Choose desired part number on the right list view, for example, lsm 6dsox, and click on select device button. In a while, Unico is configured to work with the desired adapter board. On the left toolbar, the pedometer button can be clicked to open pedometer function lab. The regression tab allows the user to start the regression. The initialization view can be used to configure the session. The session configuration includes the error metrics and the complexity level selection. The error metric is the cost function to be minimized by the tool. Several metrics are supported such as mean, mean plus standard deviation. The complexity level represents the number of iterations to be executed by the tool. Once configured the session, the user can select the folder containing the dataset and start the regression by clicking on the start button. Once finished, a folder containing the pedometer configuration and the regression report is generated under Unico root folder. This configuration contains an optimal pedometer configuration for the input dataset and can be easily loaded into the device in the user project. All the accelerometer sensor and pedometer parameters related to the embedded algorithm are summarized in the table. Some of them can be manually changed by the user. Some others can be tuned to exploiting the regression tool. Thank you for watching this video tutorial. Should you have any doubt or should you be willing to evaluate the pedometer algorithm and its performance, please feel free to get in touch with one of ST's distributors or ST's sales offices to get support.