Related Experiment Videos
Compression of surface EMG signals with algebraic code excited linear prediction.
Elias Carotti1, Juan Carlos De Martin, Roberto Merletti
1Dipartimento di Automatica e Informatica (DAUIN) - Politecnico di Torino, Torino, Italy.
Medical Engineering & Physics
|May 6, 2006
Summary
This study introduces an Algebraic Code Excited Linear Prediction (ACELP) method for compressing surface electromyographic (EMG) signals, achieving 87.3% compression with minimal data loss. The technique offers efficient, low-distortion EMG signal compression suitable for long-term recordings.
Area of Science:
- Biomedical Engineering
- Signal Processing
- Rehabilitation Technology
Background:
- Long-term surface electromyographic (EMG) signal recordings are crucial for various applications.
- Effective data compression techniques for EMG signals are limited, hindering efficient storage and transmission.
- Surface EMG signal characteristics present unique challenges for traditional compression algorithms.
Purpose of the Study:
- To investigate the efficacy of a lossy coding technique for surface EMG signals.
- To adapt the Algebraic Code Excited Linear Prediction (ACELP) paradigm, commonly used in speech coding, for EMG signal compression.
- To evaluate the performance of the proposed ACELP-based compression method in terms of compression ratio, signal reconstruction accuracy, and spectral distortion.
Main Methods:
- Adapted the Algebraic Code Excited Linear Prediction (ACELP) algorithm for surface EMG signal characteristics.
- Tested the adapted ACELP algorithm on both simulated and experimental EMG datasets.
- Evaluated compression ratio, mean square error (MSE) in reconstruction, percentage error in average rectified value (ARV), and errors in spectral parameters (mean power spectral frequency, third-order power spectral moment).
Main Results:
- Achieved a high compression ratio of 87.3% for surface EMG signals.
- Demonstrated low distortion with MSE of 6.74% and ARV error of 3.11% for experimental signals.
- Maintained spectral fidelity with relative errors below 3.74% for mean power spectral frequency and 5.95% for third-order power spectral moment on experimental signals.
Conclusions:
- The proposed ACELP-based coding scheme effectively compresses surface EMG signals at high rates with low distortion.
- The method exhibits moderate computational complexity (approx. 20 MIPS) and a short algorithmic delay (approx. 160 ms).
- This technique is suitable for applications requiring efficient storage and transmission of long-term surface EMG recordings.