PTAB

IPR2023-01284

Sonos Inc v. Google LLC

Key Events

Petition

petition

Title: Coordination Amongst Multiple Voice Interface Devices
Brief Description: The ’311 patent discloses methods for coordinating among multiple voice-activated electronic devices to select a single "leader" device responsible for responding to a user's voice input. This process, known as "arbitration," is intended to prevent user confusion from multiple devices responding simultaneously.

Prior Art Relied Upon: Piersol (Application # 2017/0357478).
Core Argument for this Ground:
- Prior Art Mapping: Petitioner argued that Piersol discloses every limitation of the challenged claims. Piersol describes a system of multiple electronic devices that detect a "spoken trigger" (e.g., "Hey Siri") from a user. In response, each device determines a value or "score" based on the audio quality (e.g., signal-to-noise ratio), broadcasts its score to the other devices, receives scores from them, and compares the scores. The device with the highest score responds to the user's command, while the other devices forgo responding, thus anticipating the claimed arbitration process.
- Key Aspects: This ground asserted that Piersol's disclosure of detecting a "spoken trigger" independent of full speech-to-text processing satisfies the patent owner's own narrow interpretation of the "detect a voice input" limitation.

Prior Art Relied Upon: Piersol (Application # 2017/0357478) and Meyers (Application # 2017/0083285).
Core Argument for this Ground:
- Prior Art Mapping: This ground was presented as an alternative in case Piersol was found not to adequately disclose the "detecting a voice input" limitation. Petitioner contended that Meyers explicitly discloses this element through its description of a "voice activity detector (VAD)" and a "wake word detector" that function to detect voice presence and keywords without performing full speech transcription. Meyers teaches using these detection methods to initiate an arbitration process based on signal attributes like signal-to-noise ratio. The remaining limitations were argued to be taught by Piersol, as detailed in Ground 1A.
- Motivation to Combine: A POSITA would combine Piersol with Meyers because both references address the identical problem of duplicate responses from multiple voice-enabled devices and disclose very similar arbitration solutions. A POSITA looking to implement Piersol's system would have been motivated to look to a reference like Meyers for specific, well-known implementation details for voice detection, such as VAD or wakeword detection.
- Expectation of Success: A POSITA would have a reasonable expectation of success, as combining Meyers's standard voice detection techniques with Piersol's similar arbitration system involved applying known methods to achieve predictable results.

Prior Art Relied Upon: Masahide (Japanese Publication No. 2003/223188).
Core Argument for this Ground:
- Prior Art Mapping: Petitioner argued that Masahide discloses a voice input system with a plurality of networked "voice input devices." Each device detects a user's voice, calculates "determinative information" such as a signal-to-noise ratio (SNR), and communicates this information to other devices on the network. The devices then compare their own SNR with the received SNR values, and only the device with the "largest" SNR proceeds to process the voice command, while other devices forgo processing. This process, Petitioner asserted, anticipates all limitations of the independent claims.
Additional Grounds: Petitioner asserted an additional obviousness challenge (Ground 2B) based on Masahide in view of Jang (Application # 2013/0073293). This ground argued that if Masahide were found not to teach "detecting a voice input," Jang explicitly supplies this element through its disclosure of a "voice recognition unit" that performs "voice activity detection."

"Detecting a Voice Input": Petitioner noted that the patent owner, Google, has previously contended this term requires identifying that an audio input includes voice through a process like "voice activity detection" or "hotword/wakeword detection," which is separate and mutually exclusive from "speech recognition." While disagreeing with this narrow interpretation, the Petitioner assumed it was correct for the purposes of the petition, as it forms a basis for the priority date challenge and is met by the asserted prior art.

Effective Priority Date: A central contention was that the ’311 patent is not entitled to its claimed priority date of October 9, 2014. Petitioner argued that the parent ’107 patent lacks written description support for Google's narrow interpretation of "detecting a voice input" as being exclusive of speech recognition. In fact, the ’107 patent's specification and prosecution history allegedly show that hotword detection is considered a form of speech recognition. Therefore, Petitioner argued the ’311 patent's effective priority date is no earlier than October 3, 2016, the filing date of the first application in the family with new matter. This later priority date makes Piersol and Meyers available as prior art.

Petitioner argued that discretionary denial would be inappropriate.
- Denial under §314(a) (Fintiv): Denial was argued to be improper because PTAB policy states it will not deny institution based on a parallel ITC proceeding. Furthermore, the parallel district court case has been stayed pending the outcome of the ITC investigation, resulting in minimal investment by the court and parties.
- Denial under §325(d): Denial was argued to be improper because the asserted prior art (Piersol, Meyers, Masahide, Jang) was not considered during prosecution. Petitioner contended the Examiner erred by overlooking this highly material prior art, which was published before the patent's effective priority date.

Petitioner requested institution of an IPR and cancellation of claims 1-3, 8-12, and 14-20 of the ’311 patent as unpatentable.