Crime Detection in Local Languages
Crime detection in Pakistan starts with the local police when they are contacted by people in need. The major drawback is that local law enforcement is not equipped with the latest equipment to be able to handle crimes even on a small scale. Problems are being dealt with using primitive methods that
2025-06-28 16:26:02 - Adil Khan
Crime Detection in Local Languages
Project Area of Specialization Artificial IntelligenceProject SummaryCrime detection in Pakistan starts with the local police when they are contacted by people in need. The major drawback is that local law enforcement is not equipped with the latest equipment to be able to handle crimes even on a small scale. Problems are being dealt with using primitive methods that have been used for decades.
Crime detection using audio data in local languages could help drastically reduce the crime rate at a local level. Not only will it enable the police to be able to track down criminals, but it can also aid in preventing crime. Predicting future events based on the information gathered can play a vital role in destabilizing crime rings at the neighborhood level.
To implement a system that can detect crime in audio data, we will need a speech recognition system. Speech recognition systems require the design and development of speech corpus, language models, and grammar specifications related to the language for which the system is to be developed. Corpus development includes the collection, careful annotation, cleaning, and verification of speech data. These resources are limited for the Urdu language hence speech recognition for the Urdu language is still at a very basic level. We have an unlimited supply of unfiltered data that can be used to train even the most complex systems. We aim to target this area and design a model that can benefit society.
This project offers the ability to distinguish between harmless conversations and meaningful intelligence so organizations such as the FIA and PTA can greatly benefit from it.
- Generate enough audio data to train our model properly.
- Accurately convert audio files into text.
- Annotate the text according to the model defined.
- Build an adequate crime detection model using appropriate algorithms for identifying some of the most prevalent crimes.
- Train and test data on the AI model.
- The first major milestone in the project will be generating enough audio data to be able to train the classification model. The audio files will need to be edited/ filtered to remove any noise and excess information. The bigger the dataset gathered, the more accurate the results.
- Next, the data will be converted into text using an API that offers the maximum level of accuracy.
- Then the data will have to be properly tagged and annotated according to the algorithm we have chosen to run. This dataset will contain slang sentences and words, which will be used to train the AI model.
- The model we design will determine the context of sentences using sentiment analysis, based on the occurrence of crime/slang words.
- Our agencies or crime detection departments will be able to detect/prevent crime with more accurate results on a smaller scale.
- They won’t need to monitor every call, only the calls which are identified by the model will need to be further evaluated for any possibility of a crime
- On the basis of specific keywords, we can track criminal activity.
- We can predict future crimes.
The front end of our project will basically be an interface that allows the user to upload audio files and will receive a rating that will determine whether the sentence spoken is in a good or bad context.
On the backend, we will be applying machine learning algorithms that will be able to perform sentiment analysis and determine the context behind the use of criminal words.
In runtime, the audio files will be converted into text using the suitable API, after which the data will be passed through the algorithm and we will determine the meaning.
Final Deliverable of the Project Software SystemCore Industry ITOther Industries Security Core Technology Artificial Intelligence(AI)Other TechnologiesSustainable Development Goals Good Health and Well-Being for People, Decent Work and Economic Growth, Industry, Innovation and Infrastructure, Sustainable Cities and Communities, Peace and Justice Strong InstitutionsRequired Resources| Item Name | Type | No. of Units | Per Unit Cost (in Rs) | Total (in Rs) |
|---|---|---|---|---|
| Total in (Rs) | 64600 | |||
| NVIDIA GeForce GTX 1060 6 GB | Equipment | 1 | 35000 | 35000 |
| Boya By-M1 Professional Collar Microphone | Equipment | 2 | 2000 | 4000 |
| Research and Implementation | Miscellaneous | 1 | 5000 | 5000 |
| Urdu and Punjabi Raw Speech Corpus | Equipment | 1 | 20000 | 20000 |
| U Shape 2 in 1 Audio Splitter Jack 3.5 mm to dual female | Equipment | 2 | 300 | 600 |