The IIS Chip Gallery
Squiphisto (2006)
by
| Application | Biomedical |
| Technology | 250 |
| Manufacturer | UMC |
| Type | Semester Thesis |
| Package | LCC84 |
| Dimensions | 2500μm x 2500μm |
| Voltage | 2.5 V |
| Clock | 75 MHz |
Wavelet algorithms with subsequent encoders are used in the newest image compression standards such as JPEG2000 including EBCOT. The set partitioning in hierarchical trees (SPIHT) algorithm belongs to the next generation of encoders for wavelet-transformed images employing more sophisticated coding than in previous JPEG standards. SPIHT utilizes inherent redundancy among wavelet coefficients and is especially suited for electrocardiogram (ECG) data and color image compression.
In this semester work, a VLSI implementation of a SPIHT algorithm modified for the compression of biomedical signals was realized (MSPIHT). The wavelet transformation is applied iteratively resulting in a transformed image of pyramidal form. The tree is traversed and the most significant bits are encoded first into an embedded bit stream.
For ECG signals, MSPIHT with compression ratios of up to 20:1 provides acceptable results for visual inspection. Further compression may lead to false diagnosis by medical doctors due to vanishing details. The ASIC runs typically 0.35 seconds to process an image of 512×512 pixels at the maximum clock frequency of 75 MHz. For real-time processing of ECG data with an average heart beat of 100 beats/min, a 100 kHz clock frequency is sufficient in low-power applications.