Bu çalışmada, Parkinson hastalarında sıklıkla görülen patolojik dinlenme tremorlerinin kompleks düzlemde adaptif tahmini gerçekleştirilmiştir. Bu kapsamda, ilk olarak anlık olarak ölçülen patolojik “sağ el ve sol el” veya “sağ bacak ve sol bacak” tremorleri kompleks düzlemde ifade edilmiş ve ardından, bu kompleks-değerli patolojik tremorler, bir-adım-ileri kesin linear (strcitly linear, SL) ve geniş linear (Widely linear, WL) tabanlı tahmin ediciler vasıtasıyla adaptif olarak tahmin edilmiştir. Burada, SL tabanlı tahmin edici, kompleks-değerli en küçük ortalama kare (Complex-valued least mean square, CLMS) algoritması ile eğitilirken, WL tabanlı tahmin edici ise artırılmış CLMS (Augmented CLMS) algoritması ile eğitilmiştir. Tahmin edicilerin başarımları, gerçek dünya verisi olan patolojik dinlenme tremorleri üzerinde mutlak hata ve tahmin kazancı açısından incelenmiştir. Yapılan benzetim sonuçları; kompleks-değerli patolojik dinlenme tremorlerinin dairesel olmayan davranış sergilediğini ve bu yüzden de WL tabanlı tahmin edicinin, SL versiyonuna kıyasla daha üstün bir başarım sergilediğini ortaya koymuştur.
In this study, an adaptive estimate of the pathological rest tremor, which is frequently seen in Parkinson patients, was made on a complex level. In this scope, the pathological "right and left hand" or "right leg and left leg" tremors first instantly measured were expressed in a complex flat and then, these complex-value pathological tremors were adjustablely predicted through one-step linear (strcitly linear, SL) and broad linear (Widely linear, WL) based predictors. Here, the SL-based predictor is trained with the complex-valued minimum average square (CLMS) algorithm, while the WL-based predictor is trained with the increased CLMS (Augmented CLMS) algorithm. The successes of the predictors have been studied in terms of absolute errors and predictive gains on pathological rest tremors, which are real world data. The comparison results showed that complex-value pathological rest tremors showed non-circular behavior and therefore the WL-based predictor showed a superior success compared to the SL version.
In this study, adaptive estimation of pathological resting tremors, which is frequently encountered in Parkinson’s patients, is performed in the complex domain. In this context, pathological “right hand and left hand” or “right leg and left leg” tremors, which were measured instantaneously, are first expressed in the complex domain. Then, these complex-valued pathological tremors are predicted adaptively using one-step-ahead strictly linear (SL) and widely linear (WL) based predictors. Here, the SL based predictor is trained by the Complex-valued least mean square (CLMS) algorithm, while the WL based predictor is trained by the augmented CLMS (ACLMS) algorithm. The performances of these predictors were examined in terms of absolute error and prediction gain on pathological resting tremors as real-world data. Simulation results reveal that complex-valued pathological resting tremors exhibit non-circular behavior and thus the WL based predictor outperforms the SL version.
Alan : Fen Bilimleri ve Matematik; Mühendislik
Dergi Türü : Uluslararası
Benzer Makaleler | Yazar | # |
---|
Makale | Yazar | # |
---|