Performance evaluation of dual vs. single lead automatic, real-world arrhythmic ECG recordings