Era I · 1870 — 1970
Mechanical.
The idea of a machine that could read predates computers by almost a century. Early pioneers built physical devices from selenium cells, spinning disks and vacuum tubes — driven mostly by the ambition of giving blind readers access to printed text.
- 1870
- Carey’s retina-inspired sensorT.D. Carey proposes a mosaic of selenium photocells that converts an image into electrical signals. Decades ahead of the available technology.
- 1885
- Nipkow’s scanning diskA rotating disk with spiral holes scans an image point-by-point into a serial electrical signal. The scanning principle persists in every OCR device for 60 years.
- 1912
- Optophone for the blindd’Albe’s device maps each printed character to a distinct musical chord. A trained reader reaches about one word per minute.
- 1914
- Goldberg’s statistical machineThe first device to recognise printed characters by comparing their photocell signature against stored templates. Ancestor of all template-based OCR.
- 1929
- Tauschek’s template patentA spinning disk with cut-out letters; maximum light transmission identifies the character. Elegant and impossibly slow.
- 1931
- IBM acquires Goldberg’s patentsThe technology sits dormant for twenty years, waiting for electronics to catch up.
- 1949
- RCA reading machineThe US Veterans Administration funds the first prototype that reads printed pages aloud. Accuracy under 50% — but OCR now has serious government funding.
- 1951
- GISMO — first electronic OCRNIST’s Sheppard replaces the spinning disk with static photocell arrays. The leap from mechanical to electronic is the most important transition in OCR history.
- 1955
- MICR for bankingThe American Bankers Association adopts the E-13B magnetic-ink font for check processing. Not optical — but it proves banks will pay for machine reading.
- 1957
- The perceptron detourRosenblatt’s Mark I Perceptron barely distinguishes triangles from squares. Minsky’s 1969 critique kills neural networks for two decades.
- 1965
- First commercial OCRReader’s Digest + RCA process 1,500 documents per hour — but only in the purpose-built OCR-A font.
- 1966
- US Postal ServiceMachine-sorting mail using OCR scanners. The first industrial-scale deployment.
- 1968
- Kurzweil’s insightExtract structural features (strokes, apexes, crossbars), then classify. The separation of feature extraction from classification is the architecture every modern OCR system still uses.