Limits...
Improving activity recognition using a wearable barometric pressure sensor in mobility-impaired stroke patients.

Massé F, Gonzenbach RR, Arami A, Paraschiv-Ionescu A, Luft AR, Aminian K - J Neuroeng Rehabil (2015)

Bottom Line: The first stage processed the events to infer activities and the second stage improved activity recognition by applying behavioral constraints.The proposed classifier essentially benefits from a better recognition of standing activity (70.3 % versus 61.5 % [FIS-IMU] and 42.5 % [EPOCH]) with 98.2 % CCR for body elevation estimation.The monitoring and recognition of daily activities in mobility-impaired stoke patients can be significantly improved using a trunk-fixed sensor that integrates BP, inertial sensors, and an event-based activity classifier.

View Article: PubMed Central - PubMed

Affiliation: Laboratory of Movement Analysis and Measurement, Ecole Polytechnique Fédérale de Lausanne, Station 11, 1015, Lausanne, Switzerland. fabien.masse@epfl.ch.

ABSTRACT

Background: Stroke survivors often suffer from mobility deficits. Current clinical evaluation methods, including questionnaires and motor function tests, cannot provide an objective measure of the patients' mobility in daily life. Physical activity performance in daily-life can be assessed using unobtrusive monitoring, for example with a single sensor module fixed on the trunk. Existing approaches based on inertial sensors have limited performance, particularly in detecting transitions between different activities and postures, due to the inherent inter-patient variability of kinematic patterns. To overcome these limitations, one possibility is to use additional information from a barometric pressure (BP) sensor.

Methods: Our study aims at integrating BP and inertial sensor data into an activity classifier in order to improve the activity (sitting, standing, walking, lying) recognition and the corresponding body elevation (during climbing stairs or when taking an elevator). Taking into account the trunk elevation changes during postural transitions (sit-to-stand, stand-to-sit), we devised an event-driven activity classifier based on fuzzy-logic. Data were acquired from 12 stroke patients with impaired mobility, using a trunk-worn inertial and BP sensor. Events, including walking and lying periods and potential postural transitions, were first extracted. These events were then fed into a double-stage hierarchical Fuzzy Inference System (H-FIS). The first stage processed the events to infer activities and the second stage improved activity recognition by applying behavioral constraints. Finally, the body elevation was estimated using a pattern-enhancing algorithm applied on BP. The patients were videotaped for reference. The performance of the algorithm was estimated using the Correct Classification Rate (CCR) and F-score. The BP-based classification approach was benchmarked against a previously-published fuzzy-logic classifier (FIS-IMU) and a conventional epoch-based classifier (EPOCH).

Results: The algorithm performance for posture/activity detection, in terms of CCR was 90.4 %, with 3.3 % and 5.6 % improvements against FIS-IMU and EPOCH, respectively. The proposed classifier essentially benefits from a better recognition of standing activity (70.3 % versus 61.5 % [FIS-IMU] and 42.5 % [EPOCH]) with 98.2 % CCR for body elevation estimation.

Conclusion: The monitoring and recognition of daily activities in mobility-impaired stoke patients can be significantly improved using a trunk-fixed sensor that integrates BP, inertial sensors, and an event-based activity classifier.

No MeSH data available.


Related in: MedlinePlus

Fuzzy inference system: weighting membership function and defuzzification. (Top) Fuzzy output for rules with two different weights W = 1 on the top left graph and W = 0.5 on the top right axis. (Bottom) Defuzzification of the H-FIS: Mean of maximum. It corresponds to the middle value of the maximum membership plateau
© Copyright Policy - OpenAccess
Related In: Results  -  Collection

License 1 - License 2
getmorefigures.php?uid=PMC4549072&req=5

Fig5: Fuzzy inference system: weighting membership function and defuzzification. (Top) Fuzzy output for rules with two different weights W = 1 on the top left graph and W = 0.5 on the top right axis. (Bottom) Defuzzification of the H-FIS: Mean of maximum. It corresponds to the middle value of the maximum membership plateau

Mentions: A fuzzy inference system is generally defined by a set of membership functions to transfer its inputs into fuzzy (linguistic) variables, a set of “If-Then” rules to fuse the fuzzy variables and map the antecedents to consequences and, an implication and aggregation operator, and finally a de-fuzzification method. As an example, the first rule from Table 2 should be read as: “If the PrevAct is sitting AND CurrAct is unknown AND Transition is Detected AND TransitionType is SiSt THEN EventActivity is standing”. As another example, a fuzzy inference system (used for classification) and its fuzzy rules are presented in Fig. 4. This fuzzy inference system works as follows. The first step is to fuzzify the input using the membership functions. In this example, we assumed following values: PrevAct = −0.75 (fell into sit, unknown and stand activities with decreasing membership values respectively); CurrAct = 0 (fell into unknown, sit and stand activities); Transition = 0.7; TransitionType = 0.6. Their degrees of membership are computed through the membership function as showed in Fig. 4 for the associated rules (e.g. TransitionType with SiSt). The corresponding degree of membership for each variable is denoted by the shaded area in each graph of the input variables. They are for example: 0.88 for the PrevAct to be sitting, 1 for CurrAct to be member of Unknown, 0.83 for the Transition to be Detected and 0.66 for TransitionType to be SiSt. To compute the contributions of each variable to the rule, an implication operator is used: the minimum computed across each variable. The result of each rule is reported as a fuzzy output with a degree of membership corresponding to this minimum value. In this example and for the first rule, the minimum is 0.66 (from TransitionType). For the second and third rules, the fuzzy output of these rules (minimum values) were also computed: 0.33 for the 2nd rule and 0.12 for the 3rd rule. The next step is to bring together the contributions from each rule: the aggregation step. In our case, the maximum operator was used to merge the contributions and a polygon shape is therefore obtained as shown in Fig. 4. The last step, call the defuzzification step, computes the output of the FIS from this polygon. For the Event-FIS, the output correspond to the (x-axis value of the) centroid of polygon. Furthermore, weights were associated with the rules to change their contributions according to the confidence level of the rule. This provides an opportunity for to favor one rule with respect to another. An example of a weighted rule output is presented in Fig. 5.Fig. 4


Improving activity recognition using a wearable barometric pressure sensor in mobility-impaired stroke patients.

Massé F, Gonzenbach RR, Arami A, Paraschiv-Ionescu A, Luft AR, Aminian K - J Neuroeng Rehabil (2015)

Fuzzy inference system: weighting membership function and defuzzification. (Top) Fuzzy output for rules with two different weights W = 1 on the top left graph and W = 0.5 on the top right axis. (Bottom) Defuzzification of the H-FIS: Mean of maximum. It corresponds to the middle value of the maximum membership plateau
© Copyright Policy - OpenAccess
Related In: Results  -  Collection

License 1 - License 2
Show All Figures
getmorefigures.php?uid=PMC4549072&req=5

Fig5: Fuzzy inference system: weighting membership function and defuzzification. (Top) Fuzzy output for rules with two different weights W = 1 on the top left graph and W = 0.5 on the top right axis. (Bottom) Defuzzification of the H-FIS: Mean of maximum. It corresponds to the middle value of the maximum membership plateau
Mentions: A fuzzy inference system is generally defined by a set of membership functions to transfer its inputs into fuzzy (linguistic) variables, a set of “If-Then” rules to fuse the fuzzy variables and map the antecedents to consequences and, an implication and aggregation operator, and finally a de-fuzzification method. As an example, the first rule from Table 2 should be read as: “If the PrevAct is sitting AND CurrAct is unknown AND Transition is Detected AND TransitionType is SiSt THEN EventActivity is standing”. As another example, a fuzzy inference system (used for classification) and its fuzzy rules are presented in Fig. 4. This fuzzy inference system works as follows. The first step is to fuzzify the input using the membership functions. In this example, we assumed following values: PrevAct = −0.75 (fell into sit, unknown and stand activities with decreasing membership values respectively); CurrAct = 0 (fell into unknown, sit and stand activities); Transition = 0.7; TransitionType = 0.6. Their degrees of membership are computed through the membership function as showed in Fig. 4 for the associated rules (e.g. TransitionType with SiSt). The corresponding degree of membership for each variable is denoted by the shaded area in each graph of the input variables. They are for example: 0.88 for the PrevAct to be sitting, 1 for CurrAct to be member of Unknown, 0.83 for the Transition to be Detected and 0.66 for TransitionType to be SiSt. To compute the contributions of each variable to the rule, an implication operator is used: the minimum computed across each variable. The result of each rule is reported as a fuzzy output with a degree of membership corresponding to this minimum value. In this example and for the first rule, the minimum is 0.66 (from TransitionType). For the second and third rules, the fuzzy output of these rules (minimum values) were also computed: 0.33 for the 2nd rule and 0.12 for the 3rd rule. The next step is to bring together the contributions from each rule: the aggregation step. In our case, the maximum operator was used to merge the contributions and a polygon shape is therefore obtained as shown in Fig. 4. The last step, call the defuzzification step, computes the output of the FIS from this polygon. For the Event-FIS, the output correspond to the (x-axis value of the) centroid of polygon. Furthermore, weights were associated with the rules to change their contributions according to the confidence level of the rule. This provides an opportunity for to favor one rule with respect to another. An example of a weighted rule output is presented in Fig. 5.Fig. 4

Bottom Line: The first stage processed the events to infer activities and the second stage improved activity recognition by applying behavioral constraints.The proposed classifier essentially benefits from a better recognition of standing activity (70.3 % versus 61.5 % [FIS-IMU] and 42.5 % [EPOCH]) with 98.2 % CCR for body elevation estimation.The monitoring and recognition of daily activities in mobility-impaired stoke patients can be significantly improved using a trunk-fixed sensor that integrates BP, inertial sensors, and an event-based activity classifier.

View Article: PubMed Central - PubMed

Affiliation: Laboratory of Movement Analysis and Measurement, Ecole Polytechnique Fédérale de Lausanne, Station 11, 1015, Lausanne, Switzerland. fabien.masse@epfl.ch.

ABSTRACT

Background: Stroke survivors often suffer from mobility deficits. Current clinical evaluation methods, including questionnaires and motor function tests, cannot provide an objective measure of the patients' mobility in daily life. Physical activity performance in daily-life can be assessed using unobtrusive monitoring, for example with a single sensor module fixed on the trunk. Existing approaches based on inertial sensors have limited performance, particularly in detecting transitions between different activities and postures, due to the inherent inter-patient variability of kinematic patterns. To overcome these limitations, one possibility is to use additional information from a barometric pressure (BP) sensor.

Methods: Our study aims at integrating BP and inertial sensor data into an activity classifier in order to improve the activity (sitting, standing, walking, lying) recognition and the corresponding body elevation (during climbing stairs or when taking an elevator). Taking into account the trunk elevation changes during postural transitions (sit-to-stand, stand-to-sit), we devised an event-driven activity classifier based on fuzzy-logic. Data were acquired from 12 stroke patients with impaired mobility, using a trunk-worn inertial and BP sensor. Events, including walking and lying periods and potential postural transitions, were first extracted. These events were then fed into a double-stage hierarchical Fuzzy Inference System (H-FIS). The first stage processed the events to infer activities and the second stage improved activity recognition by applying behavioral constraints. Finally, the body elevation was estimated using a pattern-enhancing algorithm applied on BP. The patients were videotaped for reference. The performance of the algorithm was estimated using the Correct Classification Rate (CCR) and F-score. The BP-based classification approach was benchmarked against a previously-published fuzzy-logic classifier (FIS-IMU) and a conventional epoch-based classifier (EPOCH).

Results: The algorithm performance for posture/activity detection, in terms of CCR was 90.4 %, with 3.3 % and 5.6 % improvements against FIS-IMU and EPOCH, respectively. The proposed classifier essentially benefits from a better recognition of standing activity (70.3 % versus 61.5 % [FIS-IMU] and 42.5 % [EPOCH]) with 98.2 % CCR for body elevation estimation.

Conclusion: The monitoring and recognition of daily activities in mobility-impaired stoke patients can be significantly improved using a trunk-fixed sensor that integrates BP, inertial sensors, and an event-based activity classifier.

No MeSH data available.


Related in: MedlinePlus