User Emotion Recognition Based on Multi-class Sensors of Smartphone

A Review of Emotion Recognition Methods Based on Data Acquired via Smartphone Sensors

“…A time-based schedule is most common. It may be performed several times a day [52,60,61,64,69,75,82,83], sometimes more often, e.g., 12 times a day [50] or every hour [84]. Usually some interval between subsequent reports is required [60,83].…”

“…A sequence of the acceleration values read during a time window are used to extract features. The parameter values may be calculated either on the aggregated series [110] or only on individual of the three axes [40,43,60,82] or by applying both approaches [48,63,68]. In any of these cases, the feature vectors are extracted on the basis of segmented data series, which results in obtaining one feature vector for each segment.…”

“…In any of these cases, the feature vectors are extracted on the basis of segmented data series, which results in obtaining one feature vector for each segment. The common way to split the series into segments is a sliding window, which may be either overlapping or not, and may be of different length specified by a time period, e.g., 5 s [43,74,82], or a number of samples, e.g., 128 [40]. If the sensor data is recorded for a specified period, e.g., 5 min, then it leads to a number of feature vectors being extracted.…”

“…Apart from the already mentioned skewness and kurtosis, which belong to descriptive statistics and measure the asymmetry and the heaviness of the distribution tails respectively, the waveform shape may be also described by identifying crests and troughs, which are local maxima and minima. Several features may be calculated on the basis of them, e.g., crest mean, trough mean, and the maximum and minimum difference between the crest and trough [82]. In [48], the zero crossing rate and slope sign change were also incorporated.…”

Support for Employees with ASD in the Workplace Using a Bluetooth Skin Resistance Sensor–A Preliminary Study

“…It uses information from audio, physical activity, communication data collected, during the workday, and heart rate variability data collected at night, during sleep, to build multinomial logistic regression models [ 34 ]. There are also other systems using smartphones and wireless wearable sensors for detecting human activity [ 35 , 36 , 37 ], emotion recognition and classification [ 38 , 39 ] or stress monitoring e.g., “AMMON” [ 40 ], “MoodSense” [ 41 ],“StressSense” [ 42 ] and sensors based on the Heart Rate Variability [ 43 , 44 ]. We can also find more than a thousand commercially available smartphone applications for stress recognition.…”