Recognition Accuracy

The Vestec Speech Engine delivers among the highest recognition accuracy in the industry. The engine is well suited for a variety of keywords, names, digits, numbers, dates, and yes/no grammars. In addition, the engine has been designed to operate in different channels – including VoIP, cellular, and landline – in both “noisy” as well as “clean” environments.

Generally speaking, the Vestec engine delivers speech recognition accuracy in the 90% range for native speakers and in the 80% range for non-native speakers for most applications without “tuning”. Recognition accuracy typically improves by 5-10% with grammar “tuning” by the application developer.

Accuracy Study

Vestec recently benchmarked the recognition accuracy of its speech engine against a leading competitor in the market today. The test was based on use of independent third-party audio data consisting of human recordings – in a variety of voices, in different channels – of the 500 most commonly spoken words in English. For both native and non-native speakers, Vestec’s engine outperformed the competitor by nearly 3% points in recognition accuracy.

NOTE: The above results are for "untuned" speech grammars using default engine settings. Grammar "tuning" by the application developer can result in improvement recognition accuracy.