**3.6 Speech and audio processing**

formats and produce a tool that will leverage on already available retrieval

• Person and object retrieval and identification

mobile videotechnology should be deployed.

application terms.

• People reidentification

placed in the neighborhood.

**146**

community.

*Digital Forensic Science*

the search.

should be developed:

techniques provided either by the developer of CCTV system or by the open-source

This task will be aimed to develop tools to aid the human expert looks for individuals or objects that are useful for the investigation. Object of interests can be, for example, heads, vehicles, license plates, guns, dresses and all other objects that can link a person to the event, etc. The tool will handle the cases of videos with low quality, bad lighting condition, camera/object position and facial expressions. The persons/objects with enough quality retrieved from the video will be compared by an automatic procedure to a set of known elements of comparison. The expert will then provide feedback on the comparison results in order to refine

An important main focus of police work is the identification of people for which a decision of the public prosecutor's office or a judge to the observation or an arrest warrant was issued. Within this scope of arrangement, the use of video-supervised places and facilities should be used. At earlier not known places, the application of

The aim of this task is to develop methods and procedures for an automatic system for identification of one or several target people in mobile video recordings

1.Prototype-based methods and procedures for the identification of an aim person or personal group based on one or several prototypes obtained from pictures of the target person or target personal group without that the system on several picture sequences about likelihood must be trained.

2.Methods and procedures which allow it from a picture of the person (passport photo or photo of an observation) on site or fast to generate a prototype picture, which includes aging processes of the person and different perspectives and eliminates covers of clothes must be included

3.To be able to learn methods and procedures to the generalization about case uses around the model for the facial recognition under these

This task is aimed at devising techniques for the re-identification of people in videos captured by different cameras. In many outdoor and indoor environments, different cameras are present to monitor a given area (e.g., in a given street, you can find cameras operated by shops, banks, etc., or by the municipal police). The development of reidentification techniques allows tracking a subject that exits from the field of view of a camera and enters into the field of view of another camera

without special facilities of the forensic disciplines.

based on passport photos or other available pictures. On this occasion, the prototype-based methods should be used, which are able to work on different picture representations, like pixel, features and graphs. By means of case-based learning mechanisms, a model should be automatically learnt for procedures to the facial recognition under the described application terms. In detail, the following

A significant portion the data collected by law enforcement agencies are speech, sound and audio files.

Similarity, case-based reasoning, compressive reasoning and sensing-based novel methods for speech and audio recognition using both temporal and frequency domain information will be developed. Similarity learning, case generalization and case storage and compressive learning and sensing will allow the handling of very large amount (terabytes) of data.

• Speech and audio representation

Development of speech and audio representation schemes using both spectral, wavelet scattering and temporal methods such as delta modulation and zerocrossing information. This novel representation will be used in all of the above tasks.

• Case and similarity-based reasoning-based speech and audio recognition

Development of 'query by example,' keyword and phrase-based retrieval schemes using conventional and structural similarity-based methods capable of part and whole similarity matching. Once the keyword and phrases are detected, analysts can manually process the proposed retrieval results.

• Compressive reasoning and recognition

Exemplar-based speech, speaker and audio recognition. The resulting scheme will be computationally efficient, and it will work in the compressed data domain.
