Obviousness

Obviousness Analysis of U.S. Patent 12,406,663 under 35 U.S.C. § 103

Washington, D.C. - An analysis of the claims of U.S. Patent 12,406,663 ("the '663 patent") in light of prior art available before its priority date of December 21, 2020, indicates that the claims may be rendered obvious under 35 U.S.C. § 103. The analysis centers on the combination of known elements from separate prior art references which, when combined, appear to teach all elements of the independent claims of the '663 patent. A person having ordinary skill in the art (PHOSITA) in the fields of in-vehicle infotainment, voice recognition, and IoT device control would have been motivated to combine these teachings to achieve a predictable result.

The core inventive concept of the '663 patent is a system within a vehicle that uses natural language understanding (NLU) and contextual data (e.g., vehicle location) to interpret a user's voice command and route it to the correct external smart home or IoT ecosystem. The system then uses feedback from that ecosystem to improve its NLU models.

Several combinations of prior art references could render the claims obvious. One strong combination is U.S. Patent 9,734,839 to Amazon Technologies, Inc. ("'839 patent") in view of U.S. Patent Application Publication 2014/0365228 to Honda Motor Co., Ltd. ("'228 application") and U.S. Patent Application Publication 2014/0136202 to GM Global Technology Operations LLC ("'202 application").

---

Analysis of Independent Claim 1

Claim 1 recites a system with a recognition module and a connection manager in a vehicle. The recognition module receives an utterance, uses an NLU module to interpret its meaning and identify a target smart home ecosystem, and the connection manager transmits the command. Crucially, the system receives feedback from the ecosystem and updates its NLU models.

1. Primary Reference: US 9,734,839 ('839 patent)
   The '839 patent, titled "Routing natural language commands to the appropriate applications," teaches the fundamental concept of the '663 patent. The '839 patent discloses a system that receives a natural language command, uses a routing component to determine which of several available applications (analogous to the '663 patent's "ecosystems") is the intended target, and forwards the command to that application. This directly teaches the core elements of receiving an utterance, interpreting it to identify a target, and transmitting a command to that target.

2. Secondary Reference: US 2014/0365228 ('228 application)
   The '839 patent does not explicitly place its system within a vehicle or use vehicle-specific context. The '228 application, titled "Interpretation of ambiguous vehicle instructions," remedies this. It explicitly teaches a system for interpreting voice commands within a vehicle and using contextual information, such as vehicle location ("the present location of the host vehicle"), to resolve ambiguity.

   • Motivation to Combine '839 and '228: A PHOSITA would have been motivated to implement the command-routing system of the '839 patent within the vehicle environment described in the '228 application. By 2020, integrating external applications and services into vehicle head units was a well-established trend. A PHOSITA would have seen a clear and predictable benefit in extending the in-vehicle voice assistant of the '228 application to control not just vehicle functions but also the wide array of external applications described in the '839 patent (e.g., smart home devices). This combination would allow a user to seamlessly interact with their digital life from their car, addressing a known market demand.

3. Tertiary Reference: US 2014/0136202 ('202 application)
   The combination of the '839 and '228 references teaches most of Claim 1, but lacks the explicit teaching of updating NLU models based on feedback from the target ecosystem. The '202 application, titled "Adaptation methods and systems for speech systems," supplies this missing element. It describes methods for adapting a speech processing system based on user interactions and confirmations, which serves as a form of feedback. This adaptation improves the accuracy of the system over time.

   • Motivation to Combine with '202: A PHOSITA, having combined the '839 and '228 references to create an in-vehicle command router, would naturally seek to improve its performance. The use of feedback loops and model adaptation, as taught by the '202 application, was a standard and well-known technique in machine learning and speech recognition to enhance accuracy. Applying this adaptation method to the combined system would be a predictable step to make the command routing more reliable, thus rendering the final element of Claim 1 obvious.

---

Analysis of Independent Claim 3

Claim 3 outlines a method that mirrors the system of Claim 1 but explicitly includes the step of "receiving contextual data relating to the one or more utterances."

• The combination of the '839 patent and the '228 application renders this claim obvious. The '839 patent teaches the overall method of receiving a command, determining a target ecosystem, transmitting the command, and receiving confirmation. The '228 application explicitly teaches the missing step of using contextual data, specifically vehicle location and time of day, to aid in this determination. A PHOSITA would combine these methods for the same reasons articulated for Claim 1: to create a more powerful and versatile in-vehicle assistant capable of controlling external devices, which was a clear direction of technology development.

---

Analysis of Independent Claim 12

Claim 12 describes a system with a recognition module in a vehicle that uses contextual data to identify a target non-vehicle ecosystem and a connection manager to route the command. This claim is broader than Claim 1 as it does not require the feedback-based NLU model update.

• This claim is rendered obvious by the combination of the '839 patent and the '228 application alone. The '839 patent teaches the recognition module and connection manager for routing commands to target ecosystems (applications). The '228 application teaches placing this capability within a vehicle and using contextual data to do so. The motivation to combine these references is to apply a known command-routing technique to the automotive domain to enhance functionality, which would have been a straightforward and predictable design choice for a PHOSITA at the time.