Zentian Ltd v Apple Inc
AI-generated text. This page was generated using artificial intelligence.
'Zentian Ltd. v. Apple Inc. is a Federal Circuit appeal concerning a patent for speech-recognition technology that identifies spoken words using hidden Markov models or neural networks. On September 25, 2026, the court affirmed a Patent Trial and Appeal Board decision finding all claims of U.S. Patent No. 10,062,377 unpatentable for obviousness. The opinion is nonprecedential.[1]
Background
Apple petitioned for inter partes review of the patent. The claimed system converts audio into feature vectors, compares those vectors with acoustic-model states and uses the resulting distances to identify words. The parties disputed whether the claimed “feature vector” could include a vector-quantized codeword representing quantities derived from the audio stream. Zentian argued that only quantities directly extracted from the stream qualified.[1]
Federal Circuit decision
The court held that the claim language covers both extracted and derived quantities. It agreed with the Board that codewords representing derived audio quantities fall within the claimed feature vector and affirmed the Board's obviousness determination. The ruling concerns this patent and these prior-art findings; it does not decide the validity of speech-recognition patents generally.[1]