DCT

1:19-cv-00165

Excalibur IP LLC v. Spotify USA Inc

Log In

I. Executive Summary and Procedural Information

  • Parties & Counsel:
  • Case Identification: 1:19-cv-00165, D. Del., 01/28/2019
  • Venue Allegations: Venue is alleged to be proper in the District of Delaware because Defendant Spotify USA, Inc. is a Delaware corporation. Venue against Defendant Spotify AB, a foreign entity, is alleged to be proper in any judicial district. The complaint also notes that Spotify has admitted to personal jurisdiction in this district in prior litigation.
  • Core Dispute: Plaintiff alleges that Defendant’s music streaming services infringe four patents related to audio content identification via "fingerprinting," community-based music recommendations, and relationship discovery based on user behavior.
  • Technical Context: The technologies at issue are foundational to modern digital media streaming services, enabling automated content identification and the generation of personalized user experiences, such as curated playlists.
  • Key Procedural History: The complaint notes a prior case in the District of Delaware in which Spotify admitted that it is subject to personal jurisdiction in the court. No other significant procedural events, such as prior litigation involving the patents-in-suit or IPR proceedings, are mentioned in the complaint.

Case Timeline

Date Event
1999-11-10 Priority Date for U.S. Patent No. 7,454,509
2000-05-03 Priority Date for U.S. Patent No. 8,352,331
2001-04-24 Priority Date for U.S. Patent Nos. 8,392,148 & 8,160,840
2008-11-18 U.S. Patent No. 7,454,509 Issued
2012-04-17 U.S. Patent No. 8,160,840 Issued
2013-01-08 U.S. Patent No. 8,352,331 Issued
2013-03-05 U.S. Patent No. 8,392,148 Issued
2019-01-28 Complaint Filed

II. Technology and Patent(s)-in-Suit Analysis

U.S. Patent No. 8,392,148 - Comparison of data signals using characteristic electronic thumbprints extracted therefrom

  • Patent Identification: U.S. Patent No. 8,392,148, “Comparison of data signals using characteristic electronic thumbprints extracted therefrom,” issued March 5, 2013 (Compl. ¶10).

The Invention Explained

  • Problem Addressed: The patent’s background section describes the technical difficulty of identifying and comparing digital data signals, such as audio files, due to variations in file formats and compression technologies. It notes that prior art techniques relying on attached metadata or watermarks are often incomplete, incorrect, or both (’148 Patent, col. 1:25-43; Compl. ¶16).
  • The Patented Solution: The invention proposes a method to overcome these issues by first extracting a "characteristic thumbprint" based on the content of the data signal itself. This thumbprint is then compared against a database of reference thumbprints to identify the signal. A key aspect of the solution is determining a match by finding multiple sections of the query thumbprint and a reference thumbprint that have start times differing by approximately the same amount, which allows for matching a clip to a full track (’148 Patent, Abstract; ’148 Patent, col. 2:18-22; Compl. ¶17).
  • Technical Importance: This content-based identification method provides a more robust and reliable alternative to metadata-dependent systems for managing and identifying large volumes of digital media (Compl. ¶¶16-17).

Key Claims at a Glance

  • The complaint asserts at least Claim 1 (Compl. ¶32).
  • Independent Claim 1 requires:
    • A method comprising:
    • obtaining, by a processor, a data signal;
    • determining, by the processor, a digital thumbprint of the data signal;
    • comparing, by the processor, the digital thumbprint with a plurality of reference thumbprints; and
    • determining, by the processor, a match of the digital thumbprint with at least one of the plurality of reference thumbprints when a plurality of different sections of the digital thumbprint that have start times differing by a given amount match a plurality of different sections of one or more reference thumbprints whose start times differ by approximately the given amount.
  • The complaint reserves the right to assert additional claims (Compl. ¶32).

U.S. Patent No. 8,160,840 - Comparison of data signals using characteristic electronic thumbprints extracted therefrom

  • Patent Identification: U.S. Patent No. 8,160,840, “Comparison of data signals using characteristic electronic thumbprints extracted therefrom,” issued April 17, 2012 (Compl. ¶11).

The Invention Explained

  • Problem Addressed: As with the related ’148 Patent, this invention addresses the need for a reliable method to identify data signals independent of their format or associated metadata (’840 Patent, col. 1:21-43; Compl. ¶16).
  • The Patented Solution: This patent focuses specifically on the method for extracting the digital thumbprint. The patented solution involves filtering the data signal into multiple, different frequency ranges; computing a measure of power within segments of those filtered signals; and then determining the "characteristic data" for the thumbprint based on these computed power measures. The resulting thumbprint is then written to a non-transitory storage medium (’840 Patent, Abstract). The specification describes these characteristic data as statistical metrics, such as the ratio of the standard deviation to the mean of the power values, which captures the variance of the signal within each frequency band (’840 Patent, col. 7:1-25).
  • Technical Importance: The invention provides a specific technical method for creating a compact, content-based, and format-agnostic identifier for digital media, which is a key enabling technology for large-scale media databases and identification services (Compl. ¶¶16-17).

Key Claims at a Glance

  • The complaint asserts at least Claim 1 (Compl. ¶45).
  • Independent Claim 1 requires:
    • A method for extracting a digital thumbprint from a data signal, comprising:
    • filtering the data signal to produce a plurality of filtered signals, at least two of which have a different frequency range;
    • computing a measure of power contained within at least a segment of each of a plurality of the filtered signals;
    • determining characteristic data for the data signal, by a computer processor, wherein the characteristic data is determined based on the computed measures of power; and
    • writing a digital thumbprint to a non-transitory computer-readable storage medium, the digital thumbprint comprising the determined characteristic data for the data signal.
  • The complaint reserves the right to assert additional claims (Compl. ¶45).

Multi-Patent Capsule: U.S. Patent No. 7,454,509 - Online playback system with community bias

  • Patent Identification: U.S. Patent No. 7,454,509, “Online playback system with community bias,” issued November 18, 2008 (Compl. ¶12).
  • Technology Synopsis: The patent addresses the challenge of personalizing online data streams (like internet radio) for individual users while still complying with copyright laws that restrict on-demand programming (’509 Patent, col. 2:52-57; Compl. ¶21). The solution involves creating a "community" of users with similar tastes and using the collective preferences of that community to bias the data stream for an individual user, thereby providing a more focused and enjoyable experience (’509 Patent, Abstract; Compl. ¶21).
  • Asserted Claims: The complaint asserts at least Claim 43 (Compl. ¶59).
  • Accused Features: Spotify's personalized playlist features, including Discover Weekly, Daily Mix, Release Radar, and Spotify Radio, are accused of infringing by determining communities of members and filtering content based on community preferences (Compl. ¶¶61-63).

Multi-Patent Capsule: U.S. Patent No. 8,352,331 - Relationship discovery engine

  • Patent Identification: U.S. Patent No. 8,352,331, “Relationship discovery engine,” issued January 8, 2013 (Compl. ¶13).
  • Technology Synopsis: The patent describes a method for discovering relationships among items (e.g., audio tracks, artists) and recommending them to users by observing user behavior rather than relying on explicit user input like questionnaires (’331 Patent, col. 5:55-58; Compl. ¶26). The invention involves accepting item selections from users, generating user logs, scoring those logs in response to a query, and determining related items based on the scoring of the logs, which indicates an item's relevance based on user behavior patterns (’331 Patent, Abstract; Compl. ¶27).
  • Asserted Claims: The complaint asserts at least Claim 1 (Compl. ¶72).
  • Accused Features: Spotify's personalized playlist and recommendation services (Discover Weekly, etc.) are accused of infringing by discovering relationships between items through accepting user item selections, generating user logs, and scoring those logs to determine result items (Compl. ¶¶74-79).

III. The Accused Instrumentality

Product Identification

  • The complaint accuses Spotify's music streaming platform, including its Premium Service and Ad-Supported Service (Compl. ¶¶32, 45).

Functionality and Market Context

  • The infringement allegations target core features of the Spotify service that are central to its user experience and market position.
    • Audio Identification: The complaint identifies Spotify's "Echoprint" fingerprinting technology as an infringing instrumentality (Compl. ¶34). This system is alleged to create unique identifiers for audio tracks by detecting musical "onsets," creating hashes based on the time differences between them, and matching these hashes against a reference database to identify a song (Compl. ¶13, p. 28). A visual in the complaint describes how Echoprint performs a match by "taking an unknown audio query Q and finding the corresponding track in the reference database" (Compl. ¶11).
    • Personalized Recommendations: The complaint accuses Spotify's personalized playlist features, such as Discover Weekly, Daily Mix, Release Radar, and Spotify Radio (Compl. ¶61). These features are alleged to function by analyzing the listening habits and playlists of its user base to identify users with similar tastes (Compl. ¶31). A flowchart provided in the complaint illustrates this process, starting with user listening habits and playlist creation, which develops a "taste profile" used to find new songs (Compl. ¶32). The complaint alleges Spotify derives revenue from these personalized playlist products (Compl. ¶60).

IV. Analysis of Infringement Allegations

’148 Patent Infringement Allegations

Claim Element (from Independent Claim 1) Alleged Infringing Functionality Complaint Citation Patent Citation
obtaining, by a processor, a data signal Spotify's Echoprint system obtains an "unknown audio query Q" to be identified ¶34 col. 4:5-10
determining, by the processor, a digital thumbprint of the data signal The Echoprint system performs a "fingerprinting process on the query signal, resulting in a set of offset hash pairs" ¶35, ¶11 col. 2:18-20
comparing, by the processor, the digital thumbprint with a plurality of reference thumbprints The system finds a corresponding track for the query Q by looking for identical hashes in a "reference database" of known tracks ¶36 col. 2:20-22
determining... a match... when a plurality of different sections... that have start times differing by a given amount match a plurality of different sections... whose start times differ by approximately the given amount Echoprint computes a histogram of time offset differences between matching hashes to find a "true score," allowing it to find a match even if the query is from a different section of the song with a different time offset ¶37 col. 9:18-47

Identified Points of Contention

  • Scope Questions: A central issue may be whether Spotify's "offset hash pairs" constitute "sections of the digital thumbprint" as understood in the patent. Further, the court may need to determine if calculating a "histogram of all time offset t differences" to find a "true score" meets the specific claim requirement of matching sections whose start times differ by "approximately the given amount."
  • Technical Questions: What is the technical construction of a "section" of a thumbprint in the context of the patent? The complaint's evidence is based on publicly available technical papers describing Echoprint, raising the question of whether these descriptions are sufficient to prove that the system's actual operation aligns with every element of the claim.

’840 Patent Infringement Allegations

Claim Element (from Independent Claim 1) Alleged Infringing Functionality Complaint Citation Patent Citation
filtering the data signal to produce a plurality of filtered signals, at least two of the filtered signals having a different frequency range Echoprint is alleged to perform onset detection "independently in 8 frequency bands, corresponding to the lowest 8 bands in the MPEG-Audio 32 band filterbank" ¶48 col. 5:48-52
computing a measure of power contained within at least a segment of each of a plurality of the filtered signals In each frequency band, the "magnitude of the complex band-pass signal... is compared to an exponentially-decaying threshold, and an onset recorded when the signal exceeds the threshold" ¶49 col. 7:48-55
determining characteristic data for the data signal... wherein the characteristic data... is determined based on the computed measures of power Echoprint creates hashes based on the "time difference between musical onsets in each band," where the onsets are detected based on the signal's magnitude (power) exceeding a threshold ¶50 col. 8:32-41
writing a digital thumbprint to a non-transitory computer-readable storage medium, the digital thumbprint comprising the determined characteristic data... The generated hashes are paired with the time they occur and stored in a database, with the complaint citing the use of Apache Solr to store the hash index in an "inverted index" (Compl. ¶26) ¶51 col. 2:50-53

Identified Points of Contention

  • Scope Questions: Does detecting a musical "onset" by comparing signal magnitude to a threshold constitute "computing a measure of power" as required by the claim? A court may need to decide if this binary event detection is equivalent to the quantitative power measurement (e.g., RMS) described in the patent specification.
  • Technical Questions: The claim requires that the "characteristic data" (the hash) be "determined based on the computed measures of power." The complaint alleges the hash is based on the time difference between onsets. A key technical question will be whether this time difference is considered legally "based on" the power measures that were used to identify the onsets, or if this represents a different technical process outside the claim's scope.

V. Key Claim Terms for Construction

’148 Patent

  • The Term: "a plurality of different sections of the digital thumbprint that have start times differing by a given amount"
  • Context and Importance: This term defines the core of the matching logic. The infringement analysis will depend on whether Spotify’s method of comparing discrete "offset hash pairs" and analyzing their time-offset distribution falls within the scope of this language. Practitioners may focus on this term because it appears to be the primary point of novelty in the matching process claimed by the patent.
  • Intrinsic Evidence for Interpretation:
    • Evidence for a Broader Interpretation: The specification discusses comparing a "test thumbprint" to reference thumbprints without rigidly defining "section," which could support an argument that any distinct components of the thumbprint, like hashes, qualify (’148 Patent, col. 9:18-29).
    • Evidence for a Narrower Interpretation: The detailed description of the database structure, which uses a tree search based on specific coordinates of the thumbprint vector (e.g., T[i]), could support an argument that a "section" refers to a more structured, contiguous portion of the thumbprint data rather than discrete hash values (’148 Patent, Fig. 3).

’840 Patent

  • The Term: "characteristic data... determined based on the computed measures of power"
  • Context and Importance: This term establishes the required link between the initial signal processing step (power measurement) and the final output (the thumbprint). The case may turn on whether creating a hash from the time difference between onsets is sufficiently "based on" the power measurements that initially identified those onsets.
  • Intrinsic Evidence for Interpretation:
    • Evidence for a Broader Interpretation: The summary of the invention broadly describes the thumbprint as being "based on statistics relating to the frequency information of the corresponding signal," which could encompass a variety of derived metrics, including timing data (’840 Patent, col. 2:46-49).
    • Evidence for a Narrower Interpretation: The preferred embodiment described in the specification calculates a specific statistical metric: "the ratio of the standard deviation to the mean of the power values" (’840 Patent, col. 7:1-25). A defendant might argue this disclosure limits the term "characteristic data" to such statistical summaries, excluding timing information derived from those measures.

VI. Other Allegations

  • Indirect Infringement: The complaint alleges both inducement and contributory infringement for all four patents-in-suit. The inducement allegations are based on Spotify's alleged marketing materials, technical specifications, and the provision of infringing source code that encourage use of the claimed methods (Compl. ¶¶39, 53, 66, 81). The contributory infringement allegations are based on claims that Spotify's software is a material component for practicing the patents, is not a staple article of commerce, and is especially adapted for infringement (Compl. ¶¶40, 54, 67, 82).
  • Willful Infringement: Willfulness is alleged for all asserted patents. The basis for willfulness is alleged post-suit knowledge, stating that Spotify has knowledge of the patents "by way of this complaint" and its infringement will be willful if its activities do not cease (Compl. ¶¶38, 52, 65, 80).

VII. Analyst’s Conclusion: Key Questions for the Case

  • A core issue will be one of technical and functional mapping: do the methods described in publicly available documents for Spotify's "Echoprint" and recommendation systems—which rely on concepts like "musical onsets," "offset hash pairs," and collaborative filtering of user playlists—perform the specific, multi-step functions required by the asserted claims, which are rooted in statistical signal processing and community preference filtering?
  • A key legal question will be one of definitional scope: can claim terms such as "computing a measure of power" (’840 Patent) and matching "sections... that have start times differing by a given amount" (’148 Patent) be construed broadly enough to read on the accused functionalities, or is there a fundamental mismatch in technical operation that places Spotify’s systems outside the scope of the claims?
  • A central evidentiary question will be whether the high-level technical papers, marketing materials, and diagrams cited in the complaint are sufficient to demonstrate infringement for each claim element, or if discovery of Spotify's source code and internal technical documents will be required to resolve the key technical and definitional disputes.